Hugging Face's logo

ugaoo
/
model_88cc3c46

PEFT
Safetensors
qwen2
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files
xet
 Community
1
ugaoo commited on
Commit
59565a7
·
verified ·
1 Parent(s): 9ee3e48

Upload folder using huggingface_hub

Browse files
Files changed (20) hide show
  1. README.md +138 -0
  2. adapter_config.json +1 -1
  3. adapter_model.safetensors +3 -0
  4. checkpoint-366/README.md +202 -0
  5. checkpoint-366/adapter_config.json +37 -0
  6. checkpoint-366/adapter_model.safetensors +3 -0
  7. checkpoint-366/added_tokens.json +24 -0
  8. checkpoint-366/global_step362/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt +3 -0
  9. checkpoint-366/global_step362/mp_rank_00_model_states.pt +3 -0
  10. checkpoint-366/latest +1 -0
  11. checkpoint-366/merges.txt +0 -0
  12. checkpoint-366/rng_state.pth +3 -0
  13. checkpoint-366/scheduler.pt +3 -0
  14. checkpoint-366/special_tokens_map.json +31 -0
  15. checkpoint-366/tokenizer.json +3 -0
  16. checkpoint-366/tokenizer_config.json +208 -0
  17. checkpoint-366/trainer_state.json +2595 -0
  18. checkpoint-366/training_args.bin +3 -0
  19. checkpoint-366/vocab.json +0 -0
  20. checkpoint-366/zero_to_fp32.py +604 -0
README.md ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ license: apache-2.0
4
+ base_model: Qwen/Qwen2.5-7B-Instruct
5
+ tags:
6
+ - generated_from_trainer
7
+ datasets:
8
+ - ugaoo/medmcqa30k_normal
9
+ model-index:
10
+ - name: out/Qwen_Qwen2.5_7B_Instruct_ugaoo_medmcqa30k_normal
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ [<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
18
+ <details><summary>See axolotl config</summary>
19
+
20
+ axolotl version: `0.8.0.dev0`
21
+ ```yaml
22
+ base_model: Qwen/Qwen2.5-7B-Instruct
23
+ model_type: AutoModelForCausalLM
24
+ tokenizer_type: AutoTokenizer
25
+ trust_remote_code: true
26
+
27
+ load_in_8bit: false
28
+ load_in_4bit: true
29
+ strict: false
30
+
31
+ datasets:
32
+ - path: ugaoo/medmcqa30k_normal
33
+ type: alpaca
34
+ val_set_size: 0
35
+ output_dir: ./out/Qwen_Qwen2.5_7B_Instruct_ugaoo_medmcqa30k_normal
36
+
37
+ sequence_len: 4000
38
+ sample_packing: true
39
+ pad_to_sequence_len: true
40
+
41
+ adapter: qlora
42
+ lora_r: 256
43
+ lora_alpha: 512
44
+ lora_dropout: 0.05
45
+ lora_target_linear: true
46
+ lora_target_modules:
47
+ - q_proj
48
+ - k_proj
49
+ - v_proj
50
+ - o_proj
51
+ - up_proj
52
+ - down_proj
53
+ - gate_proj
54
+
55
+ wandb_project: testsearch
56
+ wandb_entity:
57
+ wandb_watch:
58
+ wandb_name: Qwen_Qwen2.5_7B_Instruct_ugaoo_medmcqa30k_normal
59
+ wandb_log_model:
60
+
61
+ gradient_accumulation_steps: 3
62
+ micro_batch_size: 4
63
+ num_epochs: 6
64
+ optimizer: adamw_torch
65
+ lr_scheduler: cosine
66
+ learning_rate: 5e-6
67
+
68
+ train_on_inputs: false
69
+ group_by_length: false
70
+ bf16: auto
71
+ fp16: false
72
+ tf32: false
73
+
74
+ gradient_checkpointing: true
75
+ early_stopping_patience:
76
+ resume_from_checkpoint:
77
+ logging_steps: 1
78
+ xformers_attention:
79
+ flash_attention: true
80
+
81
+ warmup_steps: 100
82
+ evals_per_epoch: 6
83
+ eval_table_size:
84
+ saves_per_epoch: 1
85
+ debug:
86
+ deepspeed:
87
+ weight_decay: 0.0
88
+ fsdp:
89
+ fsdp_config:
90
+ save_total_limit: 6
91
+ ```
92
+
93
+ </details><br>
94
+
95
+ # out/Qwen_Qwen2.5_7B_Instruct_ugaoo_medmcqa30k_normal
96
+
97
+ This model is a fine-tuned version of [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) on the ugaoo/medmcqa30k_normal dataset.
98
+
99
+ ## Model description
100
+
101
+ More information needed
102
+
103
+ ## Intended uses & limitations
104
+
105
+ More information needed
106
+
107
+ ## Training and evaluation data
108
+
109
+ More information needed
110
+
111
+ ## Training procedure
112
+
113
+ ### Training hyperparameters
114
+
115
+ The following hyperparameters were used during training:
116
+ - learning_rate: 5e-06
117
+ - train_batch_size: 4
118
+ - eval_batch_size: 4
119
+ - seed: 42
120
+ - distributed_type: multi-GPU
121
+ - gradient_accumulation_steps: 3
122
+ - total_train_batch_size: 12
123
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
124
+ - lr_scheduler_type: cosine
125
+ - lr_scheduler_warmup_steps: 100
126
+ - num_epochs: 6.0
127
+
128
+ ### Training results
129
+
130
+
131
+
132
+ ### Framework versions
133
+
134
+ - PEFT 0.14.0
135
+ - Transformers 4.49.0
136
+ - Pytorch 2.5.1+cu124
137
+ - Datasets 3.2.0
138
+ - Tokenizers 0.21.0
adapter_config.json CHANGED
@@ -6,7 +6,7 @@
6
  "eva_config": null,
7
  "exclude_modules": null,
8
  "fan_in_fan_out": null,
9
- "inference_mode": false,
10
  "init_lora_weights": true,
11
  "layer_replication": null,
12
  "layers_pattern": null,
 
6
  "eva_config": null,
7
  "exclude_modules": null,
8
  "fan_in_fan_out": null,
9
+ "inference_mode": true,
10
  "init_lora_weights": true,
11
  "layer_replication": null,
12
  "layers_pattern": null,
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a75f7c6724878723f85b74604f7edbf0a9c5cfc78dc46174c0d8839306f1d488
3
+ size 1291899552
checkpoint-366/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-7B-Instruct
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 0.14.0
checkpoint-366/adapter_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "Qwen/Qwen2.5-7B-Instruct",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": null,
8
+ "fan_in_fan_out": null,
9
+ "inference_mode": true,
10
+ "init_lora_weights": true,
11
+ "layer_replication": null,
12
+ "layers_pattern": null,
13
+ "layers_to_transform": null,
14
+ "loftq_config": {},
15
+ "lora_alpha": 512,
16
+ "lora_bias": false,
17
+ "lora_dropout": 0.05,
18
+ "megatron_config": null,
19
+ "megatron_core": "megatron.core",
20
+ "modules_to_save": null,
21
+ "peft_type": "LORA",
22
+ "r": 256,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "gate_proj",
27
+ "k_proj",
28
+ "q_proj",
29
+ "o_proj",
30
+ "down_proj",
31
+ "v_proj",
32
+ "up_proj"
33
+ ],
34
+ "task_type": "CAUSAL_LM",
35
+ "use_dora": false,
36
+ "use_rslora": false
37
+ }
checkpoint-366/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a75f7c6724878723f85b74604f7edbf0a9c5cfc78dc46174c0d8839306f1d488
3
+ size 1291899552
checkpoint-366/added_tokens.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|box_end|>": 151649,
5
+ "<|box_start|>": 151648,
6
+ "<|endoftext|>": 151643,
7
+ "<|file_sep|>": 151664,
8
+ "<|fim_middle|>": 151660,
9
+ "<|fim_pad|>": 151662,
10
+ "<|fim_prefix|>": 151659,
11
+ "<|fim_suffix|>": 151661,
12
+ "<|im_end|>": 151645,
13
+ "<|im_start|>": 151644,
14
+ "<|image_pad|>": 151655,
15
+ "<|object_ref_end|>": 151647,
16
+ "<|object_ref_start|>": 151646,
17
+ "<|quad_end|>": 151651,
18
+ "<|quad_start|>": 151650,
19
+ "<|repo_name|>": 151663,
20
+ "<|video_pad|>": 151656,
21
+ "<|vision_end|>": 151653,
22
+ "<|vision_pad|>": 151654,
23
+ "<|vision_start|>": 151652
24
+ }
checkpoint-366/global_step362/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3e8717eabf2d44d487af3b2a1baca0a514b90842bb4d5c71da7c7d4c8fd6c50
3
+ size 7751120668
checkpoint-366/global_step362/mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a51af7c1bf3606b902ec6572ca6bbc934c96abcfa742f988ad9d86190426887
3
+ size 1396311192
checkpoint-366/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step362
checkpoint-366/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-366/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:529fb6e9b34a090f49bff380b7d9d6af772ecf637afb2e6f3e6887cbbb2710b8
3
+ size 14244
checkpoint-366/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40b0a2f46e82f16e42afad794a3cb1775ba3931a614ff599b3c7d284293cf063
3
+ size 1064
checkpoint-366/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|im_end|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoint-366/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
3
+ size 11421896
checkpoint-366/tokenizer_config.json ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ }
181
+ },
182
+ "additional_special_tokens": [
183
+ "<|im_start|>",
184
+ "<|im_end|>",
185
+ "<|object_ref_start|>",
186
+ "<|object_ref_end|>",
187
+ "<|box_start|>",
188
+ "<|box_end|>",
189
+ "<|quad_start|>",
190
+ "<|quad_end|>",
191
+ "<|vision_start|>",
192
+ "<|vision_end|>",
193
+ "<|vision_pad|>",
194
+ "<|image_pad|>",
195
+ "<|video_pad|>"
196
+ ],
197
+ "bos_token": null,
198
+ "chat_template": "{%- if tools %}\n {{- '<|im_start|>system\\n' }}\n {%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n {%- else %}\n {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n {%- endif %}\n {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n {%- if messages[0]['role'] == 'system' %}\n {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n {%- else %}\n {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {{- '<|im_start|>' + message.role }}\n {%- if message.content %}\n {{- '\\n' + message.content }}\n {%- endif %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '\\n<tool_call>\\n{\"name\": \"' }}\n {{- tool_call.name }}\n {{- '\", \"arguments\": ' }}\n {{- tool_call.arguments | tojson }}\n {{- '}\\n</tool_call>' }}\n {%- endfor %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- message.content }}\n {{- '\\n</tool_response>' }}\n {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
199
+ "clean_up_tokenization_spaces": false,
200
+ "eos_token": "<|im_end|>",
201
+ "errors": "replace",
202
+ "extra_special_tokens": {},
203
+ "model_max_length": 131072,
204
+ "pad_token": "<|endoftext|>",
205
+ "split_special_tokens": false,
206
+ "tokenizer_class": "Qwen2Tokenizer",
207
+ "unk_token": null
208
+ }
checkpoint-366/trainer_state.json ADDED
@@ -0,0 +1,2595 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 5.913043478260869,
5
+ "eval_steps": 500,
6
+ "global_step": 366,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.016304347826086956,
13
+ "grad_norm": 45.78702163696289,
14
+ "learning_rate": 5.0000000000000004e-08,
15
+ "loss": 4.8845,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.03260869565217391,
20
+ "grad_norm": 45.26464080810547,
21
+ "learning_rate": 1.0000000000000001e-07,
22
+ "loss": 4.8307,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.04891304347826087,
27
+ "grad_norm": 45.96015167236328,
28
+ "learning_rate": 1.5000000000000002e-07,
29
+ "loss": 4.8322,
30
+ "step": 3
31
+ },
32
+ {
33
+ "epoch": 0.06521739130434782,
34
+ "grad_norm": 45.86792755126953,
35
+ "learning_rate": 2.0000000000000002e-07,
36
+ "loss": 4.8841,
37
+ "step": 4
38
+ },
39
+ {
40
+ "epoch": 0.08152173913043478,
41
+ "grad_norm": 44.19926071166992,
42
+ "learning_rate": 2.5000000000000004e-07,
43
+ "loss": 4.8013,
44
+ "step": 5
45
+ },
46
+ {
47
+ "epoch": 0.09782608695652174,
48
+ "grad_norm": 45.76038360595703,
49
+ "learning_rate": 3.0000000000000004e-07,
50
+ "loss": 4.8112,
51
+ "step": 6
52
+ },
53
+ {
54
+ "epoch": 0.11413043478260869,
55
+ "grad_norm": 45.73483657836914,
56
+ "learning_rate": 3.5000000000000004e-07,
57
+ "loss": 4.8235,
58
+ "step": 7
59
+ },
60
+ {
61
+ "epoch": 0.13043478260869565,
62
+ "grad_norm": 46.2863655090332,
63
+ "learning_rate": 4.0000000000000003e-07,
64
+ "loss": 4.895,
65
+ "step": 8
66
+ },
67
+ {
68
+ "epoch": 0.14673913043478262,
69
+ "grad_norm": 44.06720733642578,
70
+ "learning_rate": 4.5000000000000003e-07,
71
+ "loss": 4.7696,
72
+ "step": 9
73
+ },
74
+ {
75
+ "epoch": 0.16304347826086957,
76
+ "grad_norm": 43.8859977722168,
77
+ "learning_rate": 5.000000000000001e-07,
78
+ "loss": 4.7098,
79
+ "step": 10
80
+ },
81
+ {
82
+ "epoch": 0.1793478260869565,
83
+ "grad_norm": 44.059043884277344,
84
+ "learning_rate": 5.5e-07,
85
+ "loss": 4.7519,
86
+ "step": 11
87
+ },
88
+ {
89
+ "epoch": 0.1956521739130435,
90
+ "grad_norm": 43.980201721191406,
91
+ "learning_rate": 6.000000000000001e-07,
92
+ "loss": 4.7465,
93
+ "step": 12
94
+ },
95
+ {
96
+ "epoch": 0.21195652173913043,
97
+ "grad_norm": 40.88735580444336,
98
+ "learning_rate": 6.5e-07,
99
+ "loss": 4.5416,
100
+ "step": 13
101
+ },
102
+ {
103
+ "epoch": 0.22826086956521738,
104
+ "grad_norm": 39.40138244628906,
105
+ "learning_rate": 7.000000000000001e-07,
106
+ "loss": 4.5273,
107
+ "step": 14
108
+ },
109
+ {
110
+ "epoch": 0.24456521739130435,
111
+ "grad_norm": 37.21616744995117,
112
+ "learning_rate": 7.5e-07,
113
+ "loss": 4.4206,
114
+ "step": 15
115
+ },
116
+ {
117
+ "epoch": 0.2608695652173913,
118
+ "grad_norm": 35.060447692871094,
119
+ "learning_rate": 8.000000000000001e-07,
120
+ "loss": 4.3318,
121
+ "step": 16
122
+ },
123
+ {
124
+ "epoch": 0.27717391304347827,
125
+ "grad_norm": 30.492183685302734,
126
+ "learning_rate": 8.500000000000001e-07,
127
+ "loss": 4.0841,
128
+ "step": 17
129
+ },
130
+ {
131
+ "epoch": 0.29347826086956524,
132
+ "grad_norm": 28.49239730834961,
133
+ "learning_rate": 9.000000000000001e-07,
134
+ "loss": 4.0979,
135
+ "step": 18
136
+ },
137
+ {
138
+ "epoch": 0.30978260869565216,
139
+ "grad_norm": 24.898632049560547,
140
+ "learning_rate": 9.500000000000001e-07,
141
+ "loss": 3.8752,
142
+ "step": 19
143
+ },
144
+ {
145
+ "epoch": 0.32608695652173914,
146
+ "grad_norm": 22.521434783935547,
147
+ "learning_rate": 1.0000000000000002e-06,
148
+ "loss": 3.77,
149
+ "step": 20
150
+ },
151
+ {
152
+ "epoch": 0.3423913043478261,
153
+ "grad_norm": 20.311031341552734,
154
+ "learning_rate": 1.0500000000000001e-06,
155
+ "loss": 3.5973,
156
+ "step": 21
157
+ },
158
+ {
159
+ "epoch": 0.358695652173913,
160
+ "grad_norm": 19.707839965820312,
161
+ "learning_rate": 1.1e-06,
162
+ "loss": 3.6102,
163
+ "step": 22
164
+ },
165
+ {
166
+ "epoch": 0.375,
167
+ "grad_norm": 18.431489944458008,
168
+ "learning_rate": 1.1500000000000002e-06,
169
+ "loss": 3.4147,
170
+ "step": 23
171
+ },
172
+ {
173
+ "epoch": 0.391304347826087,
174
+ "grad_norm": 17.78904151916504,
175
+ "learning_rate": 1.2000000000000002e-06,
176
+ "loss": 3.2665,
177
+ "step": 24
178
+ },
179
+ {
180
+ "epoch": 0.4076086956521739,
181
+ "grad_norm": 18.014617919921875,
182
+ "learning_rate": 1.25e-06,
183
+ "loss": 3.1941,
184
+ "step": 25
185
+ },
186
+ {
187
+ "epoch": 0.42391304347826086,
188
+ "grad_norm": 17.8990478515625,
189
+ "learning_rate": 1.3e-06,
190
+ "loss": 3.0377,
191
+ "step": 26
192
+ },
193
+ {
194
+ "epoch": 0.44021739130434784,
195
+ "grad_norm": 18.60847282409668,
196
+ "learning_rate": 1.3500000000000002e-06,
197
+ "loss": 2.8489,
198
+ "step": 27
199
+ },
200
+ {
201
+ "epoch": 0.45652173913043476,
202
+ "grad_norm": 18.562536239624023,
203
+ "learning_rate": 1.4000000000000001e-06,
204
+ "loss": 2.6229,
205
+ "step": 28
206
+ },
207
+ {
208
+ "epoch": 0.47282608695652173,
209
+ "grad_norm": 19.00884246826172,
210
+ "learning_rate": 1.45e-06,
211
+ "loss": 2.4521,
212
+ "step": 29
213
+ },
214
+ {
215
+ "epoch": 0.4891304347826087,
216
+ "grad_norm": 20.419940948486328,
217
+ "learning_rate": 1.5e-06,
218
+ "loss": 2.3907,
219
+ "step": 30
220
+ },
221
+ {
222
+ "epoch": 0.5054347826086957,
223
+ "grad_norm": 21.181713104248047,
224
+ "learning_rate": 1.5500000000000002e-06,
225
+ "loss": 2.2032,
226
+ "step": 31
227
+ },
228
+ {
229
+ "epoch": 0.5217391304347826,
230
+ "grad_norm": 20.592416763305664,
231
+ "learning_rate": 1.6000000000000001e-06,
232
+ "loss": 2.004,
233
+ "step": 32
234
+ },
235
+ {
236
+ "epoch": 0.5380434782608695,
237
+ "grad_norm": 20.40924835205078,
238
+ "learning_rate": 1.6500000000000003e-06,
239
+ "loss": 1.8373,
240
+ "step": 33
241
+ },
242
+ {
243
+ "epoch": 0.5543478260869565,
244
+ "grad_norm": 18.38524627685547,
245
+ "learning_rate": 1.7000000000000002e-06,
246
+ "loss": 1.6218,
247
+ "step": 34
248
+ },
249
+ {
250
+ "epoch": 0.5706521739130435,
251
+ "grad_norm": 18.922391891479492,
252
+ "learning_rate": 1.75e-06,
253
+ "loss": 1.4708,
254
+ "step": 35
255
+ },
256
+ {
257
+ "epoch": 0.5869565217391305,
258
+ "grad_norm": 20.200698852539062,
259
+ "learning_rate": 1.8000000000000001e-06,
260
+ "loss": 1.3044,
261
+ "step": 36
262
+ },
263
+ {
264
+ "epoch": 0.6032608695652174,
265
+ "grad_norm": 20.70162582397461,
266
+ "learning_rate": 1.85e-06,
267
+ "loss": 1.1445,
268
+ "step": 37
269
+ },
270
+ {
271
+ "epoch": 0.6195652173913043,
272
+ "grad_norm": 18.18869400024414,
273
+ "learning_rate": 1.9000000000000002e-06,
274
+ "loss": 0.9504,
275
+ "step": 38
276
+ },
277
+ {
278
+ "epoch": 0.6358695652173914,
279
+ "grad_norm": 15.211593627929688,
280
+ "learning_rate": 1.9500000000000004e-06,
281
+ "loss": 0.8083,
282
+ "step": 39
283
+ },
284
+ {
285
+ "epoch": 0.6521739130434783,
286
+ "grad_norm": 12.71890640258789,
287
+ "learning_rate": 2.0000000000000003e-06,
288
+ "loss": 0.6981,
289
+ "step": 40
290
+ },
291
+ {
292
+ "epoch": 0.6684782608695652,
293
+ "grad_norm": 10.869053840637207,
294
+ "learning_rate": 2.05e-06,
295
+ "loss": 0.6018,
296
+ "step": 41
297
+ },
298
+ {
299
+ "epoch": 0.6847826086956522,
300
+ "grad_norm": 9.5787353515625,
301
+ "learning_rate": 2.1000000000000002e-06,
302
+ "loss": 0.4899,
303
+ "step": 42
304
+ },
305
+ {
306
+ "epoch": 0.7010869565217391,
307
+ "grad_norm": 9.319293975830078,
308
+ "learning_rate": 2.15e-06,
309
+ "loss": 0.3986,
310
+ "step": 43
311
+ },
312
+ {
313
+ "epoch": 0.717391304347826,
314
+ "grad_norm": 9.404044151306152,
315
+ "learning_rate": 2.2e-06,
316
+ "loss": 0.3084,
317
+ "step": 44
318
+ },
319
+ {
320
+ "epoch": 0.7336956521739131,
321
+ "grad_norm": 7.995025634765625,
322
+ "learning_rate": 2.25e-06,
323
+ "loss": 0.2373,
324
+ "step": 45
325
+ },
326
+ {
327
+ "epoch": 0.75,
328
+ "grad_norm": 4.505397319793701,
329
+ "learning_rate": 2.3000000000000004e-06,
330
+ "loss": 0.1527,
331
+ "step": 46
332
+ },
333
+ {
334
+ "epoch": 0.7663043478260869,
335
+ "grad_norm": 2.5063579082489014,
336
+ "learning_rate": 2.35e-06,
337
+ "loss": 0.1097,
338
+ "step": 47
339
+ },
340
+ {
341
+ "epoch": 0.782608695652174,
342
+ "grad_norm": 1.5846028327941895,
343
+ "learning_rate": 2.4000000000000003e-06,
344
+ "loss": 0.102,
345
+ "step": 48
346
+ },
347
+ {
348
+ "epoch": 0.7989130434782609,
349
+ "grad_norm": 1.1286852359771729,
350
+ "learning_rate": 2.4500000000000003e-06,
351
+ "loss": 0.093,
352
+ "step": 49
353
+ },
354
+ {
355
+ "epoch": 0.8152173913043478,
356
+ "grad_norm": 1.0311343669891357,
357
+ "learning_rate": 2.5e-06,
358
+ "loss": 0.0778,
359
+ "step": 50
360
+ },
361
+ {
362
+ "epoch": 0.8315217391304348,
363
+ "grad_norm": 0.6458576917648315,
364
+ "learning_rate": 2.55e-06,
365
+ "loss": 0.0883,
366
+ "step": 51
367
+ },
368
+ {
369
+ "epoch": 0.8478260869565217,
370
+ "grad_norm": 0.727554202079773,
371
+ "learning_rate": 2.6e-06,
372
+ "loss": 0.0852,
373
+ "step": 52
374
+ },
375
+ {
376
+ "epoch": 0.8641304347826086,
377
+ "grad_norm": 0.619137167930603,
378
+ "learning_rate": 2.6500000000000005e-06,
379
+ "loss": 0.0697,
380
+ "step": 53
381
+ },
382
+ {
383
+ "epoch": 0.8804347826086957,
384
+ "grad_norm": 0.38241881132125854,
385
+ "learning_rate": 2.7000000000000004e-06,
386
+ "loss": 0.0665,
387
+ "step": 54
388
+ },
389
+ {
390
+ "epoch": 0.8967391304347826,
391
+ "grad_norm": 0.541621744632721,
392
+ "learning_rate": 2.7500000000000004e-06,
393
+ "loss": 0.0658,
394
+ "step": 55
395
+ },
396
+ {
397
+ "epoch": 0.9130434782608695,
398
+ "grad_norm": 0.3869657516479492,
399
+ "learning_rate": 2.8000000000000003e-06,
400
+ "loss": 0.0728,
401
+ "step": 56
402
+ },
403
+ {
404
+ "epoch": 0.9293478260869565,
405
+ "grad_norm": 0.3768727481365204,
406
+ "learning_rate": 2.85e-06,
407
+ "loss": 0.0741,
408
+ "step": 57
409
+ },
410
+ {
411
+ "epoch": 0.9456521739130435,
412
+ "grad_norm": 0.31400591135025024,
413
+ "learning_rate": 2.9e-06,
414
+ "loss": 0.0682,
415
+ "step": 58
416
+ },
417
+ {
418
+ "epoch": 0.9619565217391305,
419
+ "grad_norm": 0.3604981303215027,
420
+ "learning_rate": 2.95e-06,
421
+ "loss": 0.0652,
422
+ "step": 59
423
+ },
424
+ {
425
+ "epoch": 0.9782608695652174,
426
+ "grad_norm": 0.4383264482021332,
427
+ "learning_rate": 3e-06,
428
+ "loss": 0.0703,
429
+ "step": 60
430
+ },
431
+ {
432
+ "epoch": 0.9945652173913043,
433
+ "grad_norm": 0.310332328081131,
434
+ "learning_rate": 3.05e-06,
435
+ "loss": 0.0664,
436
+ "step": 61
437
+ },
438
+ {
439
+ "epoch": 1.0,
440
+ "grad_norm": 0.310332328081131,
441
+ "learning_rate": 3.1000000000000004e-06,
442
+ "loss": 0.0509,
443
+ "step": 62
444
+ },
445
+ {
446
+ "epoch": 1.016304347826087,
447
+ "grad_norm": 0.665212869644165,
448
+ "learning_rate": 3.1500000000000003e-06,
449
+ "loss": 0.0633,
450
+ "step": 63
451
+ },
452
+ {
453
+ "epoch": 1.0326086956521738,
454
+ "grad_norm": 0.3108278512954712,
455
+ "learning_rate": 3.2000000000000003e-06,
456
+ "loss": 0.0648,
457
+ "step": 64
458
+ },
459
+ {
460
+ "epoch": 1.048913043478261,
461
+ "grad_norm": 0.5986258387565613,
462
+ "learning_rate": 3.2500000000000002e-06,
463
+ "loss": 0.0769,
464
+ "step": 65
465
+ },
466
+ {
467
+ "epoch": 1.065217391304348,
468
+ "grad_norm": 0.6987417936325073,
469
+ "learning_rate": 3.3000000000000006e-06,
470
+ "loss": 0.0728,
471
+ "step": 66
472
+ },
473
+ {
474
+ "epoch": 1.0815217391304348,
475
+ "grad_norm": 0.36069774627685547,
476
+ "learning_rate": 3.3500000000000005e-06,
477
+ "loss": 0.0647,
478
+ "step": 67
479
+ },
480
+ {
481
+ "epoch": 1.0978260869565217,
482
+ "grad_norm": 0.37604954838752747,
483
+ "learning_rate": 3.4000000000000005e-06,
484
+ "loss": 0.0571,
485
+ "step": 68
486
+ },
487
+ {
488
+ "epoch": 1.1141304347826086,
489
+ "grad_norm": 0.2985791862010956,
490
+ "learning_rate": 3.45e-06,
491
+ "loss": 0.0629,
492
+ "step": 69
493
+ },
494
+ {
495
+ "epoch": 1.1304347826086956,
496
+ "grad_norm": 0.3454388380050659,
497
+ "learning_rate": 3.5e-06,
498
+ "loss": 0.0644,
499
+ "step": 70
500
+ },
501
+ {
502
+ "epoch": 1.1467391304347827,
503
+ "grad_norm": 0.3371462821960449,
504
+ "learning_rate": 3.5500000000000003e-06,
505
+ "loss": 0.0645,
506
+ "step": 71
507
+ },
508
+ {
509
+ "epoch": 1.1630434782608696,
510
+ "grad_norm": 0.27834194898605347,
511
+ "learning_rate": 3.6000000000000003e-06,
512
+ "loss": 0.0678,
513
+ "step": 72
514
+ },
515
+ {
516
+ "epoch": 1.1793478260869565,
517
+ "grad_norm": 0.26285555958747864,
518
+ "learning_rate": 3.65e-06,
519
+ "loss": 0.0586,
520
+ "step": 73
521
+ },
522
+ {
523
+ "epoch": 1.1956521739130435,
524
+ "grad_norm": 0.34152188897132874,
525
+ "learning_rate": 3.7e-06,
526
+ "loss": 0.061,
527
+ "step": 74
528
+ },
529
+ {
530
+ "epoch": 1.2119565217391304,
531
+ "grad_norm": 0.2939279079437256,
532
+ "learning_rate": 3.7500000000000005e-06,
533
+ "loss": 0.0643,
534
+ "step": 75
535
+ },
536
+ {
537
+ "epoch": 1.2282608695652173,
538
+ "grad_norm": 0.395220547914505,
539
+ "learning_rate": 3.8000000000000005e-06,
540
+ "loss": 0.0644,
541
+ "step": 76
542
+ },
543
+ {
544
+ "epoch": 1.2445652173913044,
545
+ "grad_norm": 0.29400259256362915,
546
+ "learning_rate": 3.85e-06,
547
+ "loss": 0.0562,
548
+ "step": 77
549
+ },
550
+ {
551
+ "epoch": 1.2608695652173914,
552
+ "grad_norm": 0.25938117504119873,
553
+ "learning_rate": 3.900000000000001e-06,
554
+ "loss": 0.0658,
555
+ "step": 78
556
+ },
557
+ {
558
+ "epoch": 1.2771739130434783,
559
+ "grad_norm": 0.36941587924957275,
560
+ "learning_rate": 3.95e-06,
561
+ "loss": 0.0678,
562
+ "step": 79
563
+ },
564
+ {
565
+ "epoch": 1.2934782608695652,
566
+ "grad_norm": 0.26572781801223755,
567
+ "learning_rate": 4.000000000000001e-06,
568
+ "loss": 0.0588,
569
+ "step": 80
570
+ },
571
+ {
572
+ "epoch": 1.309782608695652,
573
+ "grad_norm": 0.22964634001255035,
574
+ "learning_rate": 4.05e-06,
575
+ "loss": 0.0612,
576
+ "step": 81
577
+ },
578
+ {
579
+ "epoch": 1.3260869565217392,
580
+ "grad_norm": 0.24455289542675018,
581
+ "learning_rate": 4.1e-06,
582
+ "loss": 0.059,
583
+ "step": 82
584
+ },
585
+ {
586
+ "epoch": 1.3423913043478262,
587
+ "grad_norm": 0.3925253748893738,
588
+ "learning_rate": 4.15e-06,
589
+ "loss": 0.0648,
590
+ "step": 83
591
+ },
592
+ {
593
+ "epoch": 1.358695652173913,
594
+ "grad_norm": 0.24822917580604553,
595
+ "learning_rate": 4.2000000000000004e-06,
596
+ "loss": 0.0618,
597
+ "step": 84
598
+ },
599
+ {
600
+ "epoch": 1.375,
601
+ "grad_norm": 0.2522635757923126,
602
+ "learning_rate": 4.25e-06,
603
+ "loss": 0.0568,
604
+ "step": 85
605
+ },
606
+ {
607
+ "epoch": 1.391304347826087,
608
+ "grad_norm": 0.2465311586856842,
609
+ "learning_rate": 4.3e-06,
610
+ "loss": 0.0613,
611
+ "step": 86
612
+ },
613
+ {
614
+ "epoch": 1.4076086956521738,
615
+ "grad_norm": 0.2514893412590027,
616
+ "learning_rate": 4.350000000000001e-06,
617
+ "loss": 0.0574,
618
+ "step": 87
619
+ },
620
+ {
621
+ "epoch": 1.4239130434782608,
622
+ "grad_norm": 0.2320777177810669,
623
+ "learning_rate": 4.4e-06,
624
+ "loss": 0.0502,
625
+ "step": 88
626
+ },
627
+ {
628
+ "epoch": 1.440217391304348,
629
+ "grad_norm": 0.2494516372680664,
630
+ "learning_rate": 4.450000000000001e-06,
631
+ "loss": 0.0685,
632
+ "step": 89
633
+ },
634
+ {
635
+ "epoch": 1.4565217391304348,
636
+ "grad_norm": 0.244571715593338,
637
+ "learning_rate": 4.5e-06,
638
+ "loss": 0.0588,
639
+ "step": 90
640
+ },
641
+ {
642
+ "epoch": 1.4728260869565217,
643
+ "grad_norm": 0.22765810787677765,
644
+ "learning_rate": 4.5500000000000005e-06,
645
+ "loss": 0.0591,
646
+ "step": 91
647
+ },
648
+ {
649
+ "epoch": 1.4891304347826086,
650
+ "grad_norm": 0.2349582016468048,
651
+ "learning_rate": 4.600000000000001e-06,
652
+ "loss": 0.0572,
653
+ "step": 92
654
+ },
655
+ {
656
+ "epoch": 1.5054347826086958,
657
+ "grad_norm": 0.228654682636261,
658
+ "learning_rate": 4.65e-06,
659
+ "loss": 0.0593,
660
+ "step": 93
661
+ },
662
+ {
663
+ "epoch": 1.5217391304347827,
664
+ "grad_norm": 0.3090372681617737,
665
+ "learning_rate": 4.7e-06,
666
+ "loss": 0.0592,
667
+ "step": 94
668
+ },
669
+ {
670
+ "epoch": 1.5380434782608696,
671
+ "grad_norm": 0.2116968333721161,
672
+ "learning_rate": 4.75e-06,
673
+ "loss": 0.0575,
674
+ "step": 95
675
+ },
676
+ {
677
+ "epoch": 1.5543478260869565,
678
+ "grad_norm": 0.32294484972953796,
679
+ "learning_rate": 4.800000000000001e-06,
680
+ "loss": 0.062,
681
+ "step": 96
682
+ },
683
+ {
684
+ "epoch": 1.5706521739130435,
685
+ "grad_norm": 0.35240596532821655,
686
+ "learning_rate": 4.85e-06,
687
+ "loss": 0.0561,
688
+ "step": 97
689
+ },
690
+ {
691
+ "epoch": 1.5869565217391304,
692
+ "grad_norm": 0.24272935092449188,
693
+ "learning_rate": 4.9000000000000005e-06,
694
+ "loss": 0.0591,
695
+ "step": 98
696
+ },
697
+ {
698
+ "epoch": 1.6032608695652173,
699
+ "grad_norm": 0.26952850818634033,
700
+ "learning_rate": 4.95e-06,
701
+ "loss": 0.0557,
702
+ "step": 99
703
+ },
704
+ {
705
+ "epoch": 1.6195652173913042,
706
+ "grad_norm": 0.2775789201259613,
707
+ "learning_rate": 5e-06,
708
+ "loss": 0.0624,
709
+ "step": 100
710
+ },
711
+ {
712
+ "epoch": 1.6358695652173914,
713
+ "grad_norm": 0.2162039428949356,
714
+ "learning_rate": 4.999825642177387e-06,
715
+ "loss": 0.0582,
716
+ "step": 101
717
+ },
718
+ {
719
+ "epoch": 1.6521739130434783,
720
+ "grad_norm": 0.38436195254325867,
721
+ "learning_rate": 4.999302593030069e-06,
722
+ "loss": 0.0618,
723
+ "step": 102
724
+ },
725
+ {
726
+ "epoch": 1.6684782608695652,
727
+ "grad_norm": 0.2860764265060425,
728
+ "learning_rate": 4.998430925516213e-06,
729
+ "loss": 0.0637,
730
+ "step": 103
731
+ },
732
+ {
733
+ "epoch": 1.6847826086956523,
734
+ "grad_norm": 0.2180222123861313,
735
+ "learning_rate": 4.99721076122146e-06,
736
+ "loss": 0.0565,
737
+ "step": 104
738
+ },
739
+ {
740
+ "epoch": 1.7010869565217392,
741
+ "grad_norm": 0.30026087164878845,
742
+ "learning_rate": 4.995642270341961e-06,
743
+ "loss": 0.0615,
744
+ "step": 105
745
+ },
746
+ {
747
+ "epoch": 1.7173913043478262,
748
+ "grad_norm": 0.27657821774482727,
749
+ "learning_rate": 4.99372567166064e-06,
750
+ "loss": 0.0661,
751
+ "step": 106
752
+ },
753
+ {
754
+ "epoch": 1.733695652173913,
755
+ "grad_norm": 0.23324432969093323,
756
+ "learning_rate": 4.991461232516675e-06,
757
+ "loss": 0.0563,
758
+ "step": 107
759
+ },
760
+ {
761
+ "epoch": 1.75,
762
+ "grad_norm": 0.2670609951019287,
763
+ "learning_rate": 4.98884926876821e-06,
764
+ "loss": 0.0671,
765
+ "step": 108
766
+ },
767
+ {
768
+ "epoch": 1.766304347826087,
769
+ "grad_norm": 0.2405894696712494,
770
+ "learning_rate": 4.9858901447482924e-06,
771
+ "loss": 0.0571,
772
+ "step": 109
773
+ },
774
+ {
775
+ "epoch": 1.7826086956521738,
776
+ "grad_norm": 0.23271754384040833,
777
+ "learning_rate": 4.982584273214061e-06,
778
+ "loss": 0.0598,
779
+ "step": 110
780
+ },
781
+ {
782
+ "epoch": 1.7989130434782608,
783
+ "grad_norm": 0.21662941575050354,
784
+ "learning_rate": 4.978932115289165e-06,
785
+ "loss": 0.0562,
786
+ "step": 111
787
+ },
788
+ {
789
+ "epoch": 1.8152173913043477,
790
+ "grad_norm": 0.27656564116477966,
791
+ "learning_rate": 4.974934180399447e-06,
792
+ "loss": 0.0531,
793
+ "step": 112
794
+ },
795
+ {
796
+ "epoch": 1.8315217391304348,
797
+ "grad_norm": 0.3287908434867859,
798
+ "learning_rate": 4.970591026201884e-06,
799
+ "loss": 0.0522,
800
+ "step": 113
801
+ },
802
+ {
803
+ "epoch": 1.8478260869565217,
804
+ "grad_norm": 0.1879669576883316,
805
+ "learning_rate": 4.965903258506806e-06,
806
+ "loss": 0.0625,
807
+ "step": 114
808
+ },
809
+ {
810
+ "epoch": 1.8641304347826086,
811
+ "grad_norm": 0.2798942029476166,
812
+ "learning_rate": 4.9608715311933865e-06,
813
+ "loss": 0.0546,
814
+ "step": 115
815
+ },
816
+ {
817
+ "epoch": 1.8804347826086958,
818
+ "grad_norm": 0.31130167841911316,
819
+ "learning_rate": 4.955496546118439e-06,
820
+ "loss": 0.0576,
821
+ "step": 116
822
+ },
823
+ {
824
+ "epoch": 1.8967391304347827,
825
+ "grad_norm": 0.23934200406074524,
826
+ "learning_rate": 4.949779053018519e-06,
827
+ "loss": 0.0522,
828
+ "step": 117
829
+ },
830
+ {
831
+ "epoch": 1.9130434782608696,
832
+ "grad_norm": 0.2684226632118225,
833
+ "learning_rate": 4.943719849405347e-06,
834
+ "loss": 0.0596,
835
+ "step": 118
836
+ },
837
+ {
838
+ "epoch": 1.9293478260869565,
839
+ "grad_norm": 0.31276896595954895,
840
+ "learning_rate": 4.937319780454559e-06,
841
+ "loss": 0.0609,
842
+ "step": 119
843
+ },
844
+ {
845
+ "epoch": 1.9456521739130435,
846
+ "grad_norm": 0.2928497791290283,
847
+ "learning_rate": 4.930579738887827e-06,
848
+ "loss": 0.0515,
849
+ "step": 120
850
+ },
851
+ {
852
+ "epoch": 1.9619565217391304,
853
+ "grad_norm": 0.25487199425697327,
854
+ "learning_rate": 4.923500664848327e-06,
855
+ "loss": 0.0565,
856
+ "step": 121
857
+ },
858
+ {
859
+ "epoch": 1.9782608695652173,
860
+ "grad_norm": 0.19076956808567047,
861
+ "learning_rate": 4.9160835457696075e-06,
862
+ "loss": 0.0541,
863
+ "step": 122
864
+ },
865
+ {
866
+ "epoch": 1.9945652173913042,
867
+ "grad_norm": 0.36202409863471985,
868
+ "learning_rate": 4.9083294162378545e-06,
869
+ "loss": 0.0572,
870
+ "step": 123
871
+ },
872
+ {
873
+ "epoch": 2.0,
874
+ "grad_norm": 0.36202409863471985,
875
+ "learning_rate": 4.900239357847582e-06,
876
+ "loss": 0.0578,
877
+ "step": 124
878
+ },
879
+ {
880
+ "epoch": 2.016304347826087,
881
+ "grad_norm": 0.5418670177459717,
882
+ "learning_rate": 4.891814499050762e-06,
883
+ "loss": 0.0504,
884
+ "step": 125
885
+ },
886
+ {
887
+ "epoch": 2.032608695652174,
888
+ "grad_norm": 0.24317026138305664,
889
+ "learning_rate": 4.883056014999423e-06,
890
+ "loss": 0.0557,
891
+ "step": 126
892
+ },
893
+ {
894
+ "epoch": 2.0489130434782608,
895
+ "grad_norm": 0.23591496050357819,
896
+ "learning_rate": 4.873965127381734e-06,
897
+ "loss": 0.0567,
898
+ "step": 127
899
+ },
900
+ {
901
+ "epoch": 2.0652173913043477,
902
+ "grad_norm": 0.26839157938957214,
903
+ "learning_rate": 4.864543104251587e-06,
904
+ "loss": 0.0489,
905
+ "step": 128
906
+ },
907
+ {
908
+ "epoch": 2.0815217391304346,
909
+ "grad_norm": 0.23912706971168518,
910
+ "learning_rate": 4.854791259851735e-06,
911
+ "loss": 0.0519,
912
+ "step": 129
913
+ },
914
+ {
915
+ "epoch": 2.097826086956522,
916
+ "grad_norm": 0.22615790367126465,
917
+ "learning_rate": 4.844710954430464e-06,
918
+ "loss": 0.0533,
919
+ "step": 130
920
+ },
921
+ {
922
+ "epoch": 2.114130434782609,
923
+ "grad_norm": 0.18881212174892426,
924
+ "learning_rate": 4.834303594051854e-06,
925
+ "loss": 0.0509,
926
+ "step": 131
927
+ },
928
+ {
929
+ "epoch": 2.130434782608696,
930
+ "grad_norm": 0.20288890600204468,
931
+ "learning_rate": 4.823570630399665e-06,
932
+ "loss": 0.0515,
933
+ "step": 132
934
+ },
935
+ {
936
+ "epoch": 2.1467391304347827,
937
+ "grad_norm": 0.2495969533920288,
938
+ "learning_rate": 4.812513560574832e-06,
939
+ "loss": 0.0593,
940
+ "step": 133
941
+ },
942
+ {
943
+ "epoch": 2.1630434782608696,
944
+ "grad_norm": 0.25076112151145935,
945
+ "learning_rate": 4.8011339268866505e-06,
946
+ "loss": 0.0546,
947
+ "step": 134
948
+ },
949
+ {
950
+ "epoch": 2.1793478260869565,
951
+ "grad_norm": 0.26761433482170105,
952
+ "learning_rate": 4.789433316637644e-06,
953
+ "loss": 0.0494,
954
+ "step": 135
955
+ },
956
+ {
957
+ "epoch": 2.1956521739130435,
958
+ "grad_norm": 0.22908158600330353,
959
+ "learning_rate": 4.777413361902152e-06,
960
+ "loss": 0.0556,
961
+ "step": 136
962
+ },
963
+ {
964
+ "epoch": 2.2119565217391304,
965
+ "grad_norm": 0.24989894032478333,
966
+ "learning_rate": 4.765075739298683e-06,
967
+ "loss": 0.0457,
968
+ "step": 137
969
+ },
970
+ {
971
+ "epoch": 2.2282608695652173,
972
+ "grad_norm": 0.19970546662807465,
973
+ "learning_rate": 4.752422169756048e-06,
974
+ "loss": 0.0483,
975
+ "step": 138
976
+ },
977
+ {
978
+ "epoch": 2.244565217391304,
979
+ "grad_norm": 0.2547377347946167,
980
+ "learning_rate": 4.739454418273314e-06,
981
+ "loss": 0.0562,
982
+ "step": 139
983
+ },
984
+ {
985
+ "epoch": 2.260869565217391,
986
+ "grad_norm": 0.30050888657569885,
987
+ "learning_rate": 4.726174293673612e-06,
988
+ "loss": 0.0474,
989
+ "step": 140
990
+ },
991
+ {
992
+ "epoch": 2.2771739130434785,
993
+ "grad_norm": 0.20995593070983887,
994
+ "learning_rate": 4.712583648351827e-06,
995
+ "loss": 0.0486,
996
+ "step": 141
997
+ },
998
+ {
999
+ "epoch": 2.2934782608695654,
1000
+ "grad_norm": 0.24649159610271454,
1001
+ "learning_rate": 4.698684378016223e-06,
1002
+ "loss": 0.0521,
1003
+ "step": 142
1004
+ },
1005
+ {
1006
+ "epoch": 2.3097826086956523,
1007
+ "grad_norm": 0.23422546684741974,
1008
+ "learning_rate": 4.684478421424007e-06,
1009
+ "loss": 0.0521,
1010
+ "step": 143
1011
+ },
1012
+ {
1013
+ "epoch": 2.3260869565217392,
1014
+ "grad_norm": 0.21018289029598236,
1015
+ "learning_rate": 4.669967760110908e-06,
1016
+ "loss": 0.0503,
1017
+ "step": 144
1018
+ },
1019
+ {
1020
+ "epoch": 2.342391304347826,
1021
+ "grad_norm": 0.20515775680541992,
1022
+ "learning_rate": 4.655154418114774e-06,
1023
+ "loss": 0.0545,
1024
+ "step": 145
1025
+ },
1026
+ {
1027
+ "epoch": 2.358695652173913,
1028
+ "grad_norm": 0.2237028032541275,
1029
+ "learning_rate": 4.6400404616932505e-06,
1030
+ "loss": 0.052,
1031
+ "step": 146
1032
+ },
1033
+ {
1034
+ "epoch": 2.375,
1035
+ "grad_norm": 0.24619129300117493,
1036
+ "learning_rate": 4.624627999035564e-06,
1037
+ "loss": 0.059,
1038
+ "step": 147
1039
+ },
1040
+ {
1041
+ "epoch": 2.391304347826087,
1042
+ "grad_norm": 0.20323814451694489,
1043
+ "learning_rate": 4.608919179968457e-06,
1044
+ "loss": 0.0534,
1045
+ "step": 148
1046
+ },
1047
+ {
1048
+ "epoch": 2.407608695652174,
1049
+ "grad_norm": 0.21665000915527344,
1050
+ "learning_rate": 4.592916195656322e-06,
1051
+ "loss": 0.0505,
1052
+ "step": 149
1053
+ },
1054
+ {
1055
+ "epoch": 2.4239130434782608,
1056
+ "grad_norm": 0.21209590137004852,
1057
+ "learning_rate": 4.576621278295558e-06,
1058
+ "loss": 0.055,
1059
+ "step": 150
1060
+ },
1061
+ {
1062
+ "epoch": 2.4402173913043477,
1063
+ "grad_norm": 0.19232606887817383,
1064
+ "learning_rate": 4.5600367008032135e-06,
1065
+ "loss": 0.0533,
1066
+ "step": 151
1067
+ },
1068
+ {
1069
+ "epoch": 2.4565217391304346,
1070
+ "grad_norm": 0.22477513551712036,
1071
+ "learning_rate": 4.543164776499945e-06,
1072
+ "loss": 0.0544,
1073
+ "step": 152
1074
+ },
1075
+ {
1076
+ "epoch": 2.4728260869565215,
1077
+ "grad_norm": 0.21310777962207794,
1078
+ "learning_rate": 4.5260078587873416e-06,
1079
+ "loss": 0.0534,
1080
+ "step": 153
1081
+ },
1082
+ {
1083
+ "epoch": 2.489130434782609,
1084
+ "grad_norm": 0.2207050621509552,
1085
+ "learning_rate": 4.508568340819654e-06,
1086
+ "loss": 0.0516,
1087
+ "step": 154
1088
+ },
1089
+ {
1090
+ "epoch": 2.505434782608696,
1091
+ "grad_norm": 0.22430899739265442,
1092
+ "learning_rate": 4.490848655169986e-06,
1093
+ "loss": 0.0552,
1094
+ "step": 155
1095
+ },
1096
+ {
1097
+ "epoch": 2.5217391304347827,
1098
+ "grad_norm": 0.2665068805217743,
1099
+ "learning_rate": 4.472851273490985e-06,
1100
+ "loss": 0.0517,
1101
+ "step": 156
1102
+ },
1103
+ {
1104
+ "epoch": 2.5380434782608696,
1105
+ "grad_norm": 0.20904089510440826,
1106
+ "learning_rate": 4.454578706170075e-06,
1107
+ "loss": 0.0477,
1108
+ "step": 157
1109
+ },
1110
+ {
1111
+ "epoch": 2.5543478260869565,
1112
+ "grad_norm": 0.21974575519561768,
1113
+ "learning_rate": 4.436033501979299e-06,
1114
+ "loss": 0.0465,
1115
+ "step": 158
1116
+ },
1117
+ {
1118
+ "epoch": 2.5706521739130435,
1119
+ "grad_norm": 0.2737596333026886,
1120
+ "learning_rate": 4.417218247719794e-06,
1121
+ "loss": 0.0443,
1122
+ "step": 159
1123
+ },
1124
+ {
1125
+ "epoch": 2.5869565217391304,
1126
+ "grad_norm": 0.19713006913661957,
1127
+ "learning_rate": 4.398135567860972e-06,
1128
+ "loss": 0.0565,
1129
+ "step": 160
1130
+ },
1131
+ {
1132
+ "epoch": 2.6032608695652173,
1133
+ "grad_norm": 0.24970802664756775,
1134
+ "learning_rate": 4.378788124174441e-06,
1135
+ "loss": 0.0473,
1136
+ "step": 161
1137
+ },
1138
+ {
1139
+ "epoch": 2.619565217391304,
1140
+ "grad_norm": 0.20846787095069885,
1141
+ "learning_rate": 4.359178615362725e-06,
1142
+ "loss": 0.0511,
1143
+ "step": 162
1144
+ },
1145
+ {
1146
+ "epoch": 2.6358695652173916,
1147
+ "grad_norm": 0.33354219794273376,
1148
+ "learning_rate": 4.33930977668283e-06,
1149
+ "loss": 0.0541,
1150
+ "step": 163
1151
+ },
1152
+ {
1153
+ "epoch": 2.6521739130434785,
1154
+ "grad_norm": 0.23585955798625946,
1155
+ "learning_rate": 4.319184379564716e-06,
1156
+ "loss": 0.0534,
1157
+ "step": 164
1158
+ },
1159
+ {
1160
+ "epoch": 2.6684782608695654,
1161
+ "grad_norm": 0.26923519372940063,
1162
+ "learning_rate": 4.298805231224721e-06,
1163
+ "loss": 0.052,
1164
+ "step": 165
1165
+ },
1166
+ {
1167
+ "epoch": 2.6847826086956523,
1168
+ "grad_norm": 0.20466569066047668,
1169
+ "learning_rate": 4.278175174273989e-06,
1170
+ "loss": 0.0532,
1171
+ "step": 166
1172
+ },
1173
+ {
1174
+ "epoch": 2.7010869565217392,
1175
+ "grad_norm": 0.27908194065093994,
1176
+ "learning_rate": 4.257297086321967e-06,
1177
+ "loss": 0.0497,
1178
+ "step": 167
1179
+ },
1180
+ {
1181
+ "epoch": 2.717391304347826,
1182
+ "grad_norm": 0.29970937967300415,
1183
+ "learning_rate": 4.236173879575022e-06,
1184
+ "loss": 0.0502,
1185
+ "step": 168
1186
+ },
1187
+ {
1188
+ "epoch": 2.733695652173913,
1189
+ "grad_norm": 0.2322796732187271,
1190
+ "learning_rate": 4.2148085004302205e-06,
1191
+ "loss": 0.054,
1192
+ "step": 169
1193
+ },
1194
+ {
1195
+ "epoch": 2.75,
1196
+ "grad_norm": 0.1984764039516449,
1197
+ "learning_rate": 4.1932039290643534e-06,
1198
+ "loss": 0.0491,
1199
+ "step": 170
1200
+ },
1201
+ {
1202
+ "epoch": 2.766304347826087,
1203
+ "grad_norm": 0.20822829008102417,
1204
+ "learning_rate": 4.1713631790182366e-06,
1205
+ "loss": 0.0555,
1206
+ "step": 171
1207
+ },
1208
+ {
1209
+ "epoch": 2.782608695652174,
1210
+ "grad_norm": 0.2537862956523895,
1211
+ "learning_rate": 4.149289296776369e-06,
1212
+ "loss": 0.0479,
1213
+ "step": 172
1214
+ },
1215
+ {
1216
+ "epoch": 2.7989130434782608,
1217
+ "grad_norm": 0.20386171340942383,
1218
+ "learning_rate": 4.126985361341984e-06,
1219
+ "loss": 0.049,
1220
+ "step": 173
1221
+ },
1222
+ {
1223
+ "epoch": 2.8152173913043477,
1224
+ "grad_norm": 0.28424644470214844,
1225
+ "learning_rate": 4.104454483807579e-06,
1226
+ "loss": 0.0481,
1227
+ "step": 174
1228
+ },
1229
+ {
1230
+ "epoch": 2.8315217391304346,
1231
+ "grad_norm": 0.2668056786060333,
1232
+ "learning_rate": 4.0816998069209516e-06,
1233
+ "loss": 0.0528,
1234
+ "step": 175
1235
+ },
1236
+ {
1237
+ "epoch": 2.8478260869565215,
1238
+ "grad_norm": 0.2771640419960022,
1239
+ "learning_rate": 4.058724504646834e-06,
1240
+ "loss": 0.0673,
1241
+ "step": 176
1242
+ },
1243
+ {
1244
+ "epoch": 2.8641304347826084,
1245
+ "grad_norm": 0.26568353176116943,
1246
+ "learning_rate": 4.0355317817241705e-06,
1247
+ "loss": 0.0438,
1248
+ "step": 177
1249
+ },
1250
+ {
1251
+ "epoch": 2.880434782608696,
1252
+ "grad_norm": 0.21058987081050873,
1253
+ "learning_rate": 4.012124873219094e-06,
1254
+ "loss": 0.0499,
1255
+ "step": 178
1256
+ },
1257
+ {
1258
+ "epoch": 2.8967391304347827,
1259
+ "grad_norm": 0.27742889523506165,
1260
+ "learning_rate": 3.988507044073687e-06,
1261
+ "loss": 0.047,
1262
+ "step": 179
1263
+ },
1264
+ {
1265
+ "epoch": 2.9130434782608696,
1266
+ "grad_norm": 0.21633900701999664,
1267
+ "learning_rate": 3.964681588650562e-06,
1268
+ "loss": 0.0485,
1269
+ "step": 180
1270
+ },
1271
+ {
1272
+ "epoch": 2.9293478260869565,
1273
+ "grad_norm": 0.21723268926143646,
1274
+ "learning_rate": 3.940651830273342e-06,
1275
+ "loss": 0.0535,
1276
+ "step": 181
1277
+ },
1278
+ {
1279
+ "epoch": 2.9456521739130435,
1280
+ "grad_norm": 0.21648381650447845,
1281
+ "learning_rate": 3.916421120763106e-06,
1282
+ "loss": 0.0469,
1283
+ "step": 182
1284
+ },
1285
+ {
1286
+ "epoch": 2.9619565217391304,
1287
+ "grad_norm": 0.2256888449192047,
1288
+ "learning_rate": 3.891992839970855e-06,
1289
+ "loss": 0.0469,
1290
+ "step": 183
1291
+ },
1292
+ {
1293
+ "epoch": 2.9782608695652173,
1294
+ "grad_norm": 0.3136005699634552,
1295
+ "learning_rate": 3.8673703953060685e-06,
1296
+ "loss": 0.0517,
1297
+ "step": 184
1298
+ },
1299
+ {
1300
+ "epoch": 2.994565217391304,
1301
+ "grad_norm": 0.2358432561159134,
1302
+ "learning_rate": 3.8425572212614155e-06,
1303
+ "loss": 0.0511,
1304
+ "step": 185
1305
+ },
1306
+ {
1307
+ "epoch": 3.0,
1308
+ "grad_norm": 0.4989979565143585,
1309
+ "learning_rate": 3.817556778933697e-06,
1310
+ "loss": 0.0502,
1311
+ "step": 186
1312
+ },
1313
+ {
1314
+ "epoch": 3.016304347826087,
1315
+ "grad_norm": 0.24074898660182953,
1316
+ "learning_rate": 3.792372555541064e-06,
1317
+ "loss": 0.0446,
1318
+ "step": 187
1319
+ },
1320
+ {
1321
+ "epoch": 3.032608695652174,
1322
+ "grad_norm": 0.23359227180480957,
1323
+ "learning_rate": 3.7670080639366e-06,
1324
+ "loss": 0.0444,
1325
+ "step": 188
1326
+ },
1327
+ {
1328
+ "epoch": 3.0489130434782608,
1329
+ "grad_norm": 0.21096405386924744,
1330
+ "learning_rate": 3.741466842118327e-06,
1331
+ "loss": 0.0492,
1332
+ "step": 189
1333
+ },
1334
+ {
1335
+ "epoch": 3.0652173913043477,
1336
+ "grad_norm": 0.18678264319896698,
1337
+ "learning_rate": 3.7157524527357036e-06,
1338
+ "loss": 0.0458,
1339
+ "step": 190
1340
+ },
1341
+ {
1342
+ "epoch": 3.0815217391304346,
1343
+ "grad_norm": 0.21836838126182556,
1344
+ "learning_rate": 3.6898684825926845e-06,
1345
+ "loss": 0.0471,
1346
+ "step": 191
1347
+ },
1348
+ {
1349
+ "epoch": 3.097826086956522,
1350
+ "grad_norm": 0.21061812341213226,
1351
+ "learning_rate": 3.663818542147409e-06,
1352
+ "loss": 0.0469,
1353
+ "step": 192
1354
+ },
1355
+ {
1356
+ "epoch": 3.114130434782609,
1357
+ "grad_norm": 0.2022838145494461,
1358
+ "learning_rate": 3.6376062650085918e-06,
1359
+ "loss": 0.0457,
1360
+ "step": 193
1361
+ },
1362
+ {
1363
+ "epoch": 3.130434782608696,
1364
+ "grad_norm": 0.21767382323741913,
1365
+ "learning_rate": 3.61123530742869e-06,
1366
+ "loss": 0.0468,
1367
+ "step": 194
1368
+ },
1369
+ {
1370
+ "epoch": 3.1467391304347827,
1371
+ "grad_norm": 0.2330484390258789,
1372
+ "learning_rate": 3.5847093477938955e-06,
1373
+ "loss": 0.0437,
1374
+ "step": 195
1375
+ },
1376
+ {
1377
+ "epoch": 3.1630434782608696,
1378
+ "grad_norm": 0.21127860248088837,
1379
+ "learning_rate": 3.5580320861110627e-06,
1380
+ "loss": 0.0459,
1381
+ "step": 196
1382
+ },
1383
+ {
1384
+ "epoch": 3.1793478260869565,
1385
+ "grad_norm": 0.20847401022911072,
1386
+ "learning_rate": 3.5312072434915983e-06,
1387
+ "loss": 0.0451,
1388
+ "step": 197
1389
+ },
1390
+ {
1391
+ "epoch": 3.1956521739130435,
1392
+ "grad_norm": 0.20767691731452942,
1393
+ "learning_rate": 3.5042385616324243e-06,
1394
+ "loss": 0.0434,
1395
+ "step": 198
1396
+ },
1397
+ {
1398
+ "epoch": 3.2119565217391304,
1399
+ "grad_norm": 0.21209655702114105,
1400
+ "learning_rate": 3.477129802294057e-06,
1401
+ "loss": 0.0418,
1402
+ "step": 199
1403
+ },
1404
+ {
1405
+ "epoch": 3.2282608695652173,
1406
+ "grad_norm": 0.21463370323181152,
1407
+ "learning_rate": 3.4498847467759e-06,
1408
+ "loss": 0.0408,
1409
+ "step": 200
1410
+ },
1411
+ {
1412
+ "epoch": 3.244565217391304,
1413
+ "grad_norm": 0.19667920470237732,
1414
+ "learning_rate": 3.4225071953887977e-06,
1415
+ "loss": 0.0476,
1416
+ "step": 201
1417
+ },
1418
+ {
1419
+ "epoch": 3.260869565217391,
1420
+ "grad_norm": 0.2126225084066391,
1421
+ "learning_rate": 3.3950009669249502e-06,
1422
+ "loss": 0.0438,
1423
+ "step": 202
1424
+ },
1425
+ {
1426
+ "epoch": 3.2771739130434785,
1427
+ "grad_norm": 0.21141058206558228,
1428
+ "learning_rate": 3.3673698981252385e-06,
1429
+ "loss": 0.0443,
1430
+ "step": 203
1431
+ },
1432
+ {
1433
+ "epoch": 3.2934782608695654,
1434
+ "grad_norm": 0.3029347360134125,
1435
+ "learning_rate": 3.3396178431440572e-06,
1436
+ "loss": 0.0549,
1437
+ "step": 204
1438
+ },
1439
+ {
1440
+ "epoch": 3.3097826086956523,
1441
+ "grad_norm": 0.20408247411251068,
1442
+ "learning_rate": 3.3117486730117092e-06,
1443
+ "loss": 0.0441,
1444
+ "step": 205
1445
+ },
1446
+ {
1447
+ "epoch": 3.3260869565217392,
1448
+ "grad_norm": 0.19485828280448914,
1449
+ "learning_rate": 3.283766275094454e-06,
1450
+ "loss": 0.0484,
1451
+ "step": 206
1452
+ },
1453
+ {
1454
+ "epoch": 3.342391304347826,
1455
+ "grad_norm": 0.23771652579307556,
1456
+ "learning_rate": 3.255674552552267e-06,
1457
+ "loss": 0.0478,
1458
+ "step": 207
1459
+ },
1460
+ {
1461
+ "epoch": 3.358695652173913,
1462
+ "grad_norm": 0.21850208938121796,
1463
+ "learning_rate": 3.227477423794412e-06,
1464
+ "loss": 0.0487,
1465
+ "step": 208
1466
+ },
1467
+ {
1468
+ "epoch": 3.375,
1469
+ "grad_norm": 0.23703360557556152,
1470
+ "learning_rate": 3.1991788219328657e-06,
1471
+ "loss": 0.0505,
1472
+ "step": 209
1473
+ },
1474
+ {
1475
+ "epoch": 3.391304347826087,
1476
+ "grad_norm": 0.21323414146900177,
1477
+ "learning_rate": 3.1707826942337124e-06,
1478
+ "loss": 0.0398,
1479
+ "step": 210
1480
+ },
1481
+ {
1482
+ "epoch": 3.407608695652174,
1483
+ "grad_norm": 0.28237420320510864,
1484
+ "learning_rate": 3.142293001566548e-06,
1485
+ "loss": 0.0429,
1486
+ "step": 211
1487
+ },
1488
+ {
1489
+ "epoch": 3.4239130434782608,
1490
+ "grad_norm": 0.24118061363697052,
1491
+ "learning_rate": 3.1137137178519983e-06,
1492
+ "loss": 0.0441,
1493
+ "step": 212
1494
+ },
1495
+ {
1496
+ "epoch": 3.4402173913043477,
1497
+ "grad_norm": 0.20298974215984344,
1498
+ "learning_rate": 3.085048829507406e-06,
1499
+ "loss": 0.0429,
1500
+ "step": 213
1501
+ },
1502
+ {
1503
+ "epoch": 3.4565217391304346,
1504
+ "grad_norm": 0.2397722750902176,
1505
+ "learning_rate": 3.056302334890786e-06,
1506
+ "loss": 0.0487,
1507
+ "step": 214
1508
+ },
1509
+ {
1510
+ "epoch": 3.4728260869565215,
1511
+ "grad_norm": 0.24425837397575378,
1512
+ "learning_rate": 3.027478243743106e-06,
1513
+ "loss": 0.0465,
1514
+ "step": 215
1515
+ },
1516
+ {
1517
+ "epoch": 3.489130434782609,
1518
+ "grad_norm": 0.2465757578611374,
1519
+ "learning_rate": 2.9985805766289815e-06,
1520
+ "loss": 0.0415,
1521
+ "step": 216
1522
+ },
1523
+ {
1524
+ "epoch": 3.505434782608696,
1525
+ "grad_norm": 0.2629953920841217,
1526
+ "learning_rate": 2.9696133643758663e-06,
1527
+ "loss": 0.0456,
1528
+ "step": 217
1529
+ },
1530
+ {
1531
+ "epoch": 3.5217391304347827,
1532
+ "grad_norm": 0.23994584381580353,
1533
+ "learning_rate": 2.940580647511805e-06,
1534
+ "loss": 0.0467,
1535
+ "step": 218
1536
+ },
1537
+ {
1538
+ "epoch": 3.5380434782608696,
1539
+ "grad_norm": 0.29615387320518494,
1540
+ "learning_rate": 2.911486475701835e-06,
1541
+ "loss": 0.0468,
1542
+ "step": 219
1543
+ },
1544
+ {
1545
+ "epoch": 3.5543478260869565,
1546
+ "grad_norm": 0.28419968485832214,
1547
+ "learning_rate": 2.8823349071831154e-06,
1548
+ "loss": 0.0479,
1549
+ "step": 220
1550
+ },
1551
+ {
1552
+ "epoch": 3.5706521739130435,
1553
+ "grad_norm": 0.2224627584218979,
1554
+ "learning_rate": 2.853130008198855e-06,
1555
+ "loss": 0.0436,
1556
+ "step": 221
1557
+ },
1558
+ {
1559
+ "epoch": 3.5869565217391304,
1560
+ "grad_norm": 0.2682191729545593,
1561
+ "learning_rate": 2.8238758524311316e-06,
1562
+ "loss": 0.0439,
1563
+ "step": 222
1564
+ },
1565
+ {
1566
+ "epoch": 3.6032608695652173,
1567
+ "grad_norm": 0.2094065397977829,
1568
+ "learning_rate": 2.7945765204326664e-06,
1569
+ "loss": 0.0472,
1570
+ "step": 223
1571
+ },
1572
+ {
1573
+ "epoch": 3.619565217391304,
1574
+ "grad_norm": 0.22561469674110413,
1575
+ "learning_rate": 2.7652360990576457e-06,
1576
+ "loss": 0.0426,
1577
+ "step": 224
1578
+ },
1579
+ {
1580
+ "epoch": 3.6358695652173916,
1581
+ "grad_norm": 0.1920589804649353,
1582
+ "learning_rate": 2.735858680891656e-06,
1583
+ "loss": 0.0465,
1584
+ "step": 225
1585
+ },
1586
+ {
1587
+ "epoch": 3.6521739130434785,
1588
+ "grad_norm": 0.2274886816740036,
1589
+ "learning_rate": 2.7064483636808314e-06,
1590
+ "loss": 0.0471,
1591
+ "step": 226
1592
+ },
1593
+ {
1594
+ "epoch": 3.6684782608695654,
1595
+ "grad_norm": 0.27154678106307983,
1596
+ "learning_rate": 2.677009249760268e-06,
1597
+ "loss": 0.0537,
1598
+ "step": 227
1599
+ },
1600
+ {
1601
+ "epoch": 3.6847826086956523,
1602
+ "grad_norm": 0.19993992149829865,
1603
+ "learning_rate": 2.6475454454818072e-06,
1604
+ "loss": 0.0423,
1605
+ "step": 228
1606
+ },
1607
+ {
1608
+ "epoch": 3.7010869565217392,
1609
+ "grad_norm": 0.26778337359428406,
1610
+ "learning_rate": 2.6180610606412587e-06,
1611
+ "loss": 0.0502,
1612
+ "step": 229
1613
+ },
1614
+ {
1615
+ "epoch": 3.717391304347826,
1616
+ "grad_norm": 0.22767065465450287,
1617
+ "learning_rate": 2.5885602079051354e-06,
1618
+ "loss": 0.0447,
1619
+ "step": 230
1620
+ },
1621
+ {
1622
+ "epoch": 3.733695652173913,
1623
+ "grad_norm": 0.2405000478029251,
1624
+ "learning_rate": 2.559047002236995e-06,
1625
+ "loss": 0.0463,
1626
+ "step": 231
1627
+ },
1628
+ {
1629
+ "epoch": 3.75,
1630
+ "grad_norm": 0.23877452313899994,
1631
+ "learning_rate": 2.529525560323462e-06,
1632
+ "loss": 0.0472,
1633
+ "step": 232
1634
+ },
1635
+ {
1636
+ "epoch": 3.766304347826087,
1637
+ "grad_norm": 0.2590119540691376,
1638
+ "learning_rate": 2.5e-06,
1639
+ "loss": 0.0474,
1640
+ "step": 233
1641
+ },
1642
+ {
1643
+ "epoch": 3.782608695652174,
1644
+ "grad_norm": 0.2477138489484787,
1645
+ "learning_rate": 2.470474439676539e-06,
1646
+ "loss": 0.0461,
1647
+ "step": 234
1648
+ },
1649
+ {
1650
+ "epoch": 3.7989130434782608,
1651
+ "grad_norm": 0.23399893939495087,
1652
+ "learning_rate": 2.4409529977630052e-06,
1653
+ "loss": 0.0446,
1654
+ "step": 235
1655
+ },
1656
+ {
1657
+ "epoch": 3.8152173913043477,
1658
+ "grad_norm": 0.24198178946971893,
1659
+ "learning_rate": 2.411439792094866e-06,
1660
+ "loss": 0.0482,
1661
+ "step": 236
1662
+ },
1663
+ {
1664
+ "epoch": 3.8315217391304346,
1665
+ "grad_norm": 0.22425131499767303,
1666
+ "learning_rate": 2.381938939358742e-06,
1667
+ "loss": 0.0477,
1668
+ "step": 237
1669
+ },
1670
+ {
1671
+ "epoch": 3.8478260869565215,
1672
+ "grad_norm": 0.21231728792190552,
1673
+ "learning_rate": 2.3524545545181936e-06,
1674
+ "loss": 0.0437,
1675
+ "step": 238
1676
+ },
1677
+ {
1678
+ "epoch": 3.8641304347826084,
1679
+ "grad_norm": 0.24554376304149628,
1680
+ "learning_rate": 2.322990750239733e-06,
1681
+ "loss": 0.0408,
1682
+ "step": 239
1683
+ },
1684
+ {
1685
+ "epoch": 3.880434782608696,
1686
+ "grad_norm": 0.20181581377983093,
1687
+ "learning_rate": 2.2935516363191695e-06,
1688
+ "loss": 0.0465,
1689
+ "step": 240
1690
+ },
1691
+ {
1692
+ "epoch": 3.8967391304347827,
1693
+ "grad_norm": 0.23197294771671295,
1694
+ "learning_rate": 2.2641413191083445e-06,
1695
+ "loss": 0.0437,
1696
+ "step": 241
1697
+ },
1698
+ {
1699
+ "epoch": 3.9130434782608696,
1700
+ "grad_norm": 0.18796613812446594,
1701
+ "learning_rate": 2.234763900942355e-06,
1702
+ "loss": 0.0427,
1703
+ "step": 242
1704
+ },
1705
+ {
1706
+ "epoch": 3.9293478260869565,
1707
+ "grad_norm": 0.2553354799747467,
1708
+ "learning_rate": 2.2054234795673336e-06,
1709
+ "loss": 0.0526,
1710
+ "step": 243
1711
+ },
1712
+ {
1713
+ "epoch": 3.9456521739130435,
1714
+ "grad_norm": 0.20340357720851898,
1715
+ "learning_rate": 2.1761241475688697e-06,
1716
+ "loss": 0.0405,
1717
+ "step": 244
1718
+ },
1719
+ {
1720
+ "epoch": 3.9619565217391304,
1721
+ "grad_norm": 0.2950204908847809,
1722
+ "learning_rate": 2.146869991801146e-06,
1723
+ "loss": 0.0469,
1724
+ "step": 245
1725
+ },
1726
+ {
1727
+ "epoch": 3.9782608695652173,
1728
+ "grad_norm": 0.226835697889328,
1729
+ "learning_rate": 2.117665092816885e-06,
1730
+ "loss": 0.0489,
1731
+ "step": 246
1732
+ },
1733
+ {
1734
+ "epoch": 3.994565217391304,
1735
+ "grad_norm": 0.2127283215522766,
1736
+ "learning_rate": 2.088513524298165e-06,
1737
+ "loss": 0.0444,
1738
+ "step": 247
1739
+ },
1740
+ {
1741
+ "epoch": 4.0,
1742
+ "grad_norm": 0.2127283215522766,
1743
+ "learning_rate": 2.059419352488196e-06,
1744
+ "loss": 0.0514,
1745
+ "step": 248
1746
+ },
1747
+ {
1748
+ "epoch": 4.016304347826087,
1749
+ "grad_norm": 0.5015725493431091,
1750
+ "learning_rate": 2.030386635624135e-06,
1751
+ "loss": 0.0402,
1752
+ "step": 249
1753
+ },
1754
+ {
1755
+ "epoch": 4.032608695652174,
1756
+ "grad_norm": 0.21895018219947815,
1757
+ "learning_rate": 2.0014194233710193e-06,
1758
+ "loss": 0.0427,
1759
+ "step": 250
1760
+ },
1761
+ {
1762
+ "epoch": 4.048913043478261,
1763
+ "grad_norm": 0.2607857584953308,
1764
+ "learning_rate": 1.972521756256895e-06,
1765
+ "loss": 0.0352,
1766
+ "step": 251
1767
+ },
1768
+ {
1769
+ "epoch": 4.065217391304348,
1770
+ "grad_norm": 0.2097238153219223,
1771
+ "learning_rate": 1.9436976651092143e-06,
1772
+ "loss": 0.0467,
1773
+ "step": 252
1774
+ },
1775
+ {
1776
+ "epoch": 4.081521739130435,
1777
+ "grad_norm": 0.21395832300186157,
1778
+ "learning_rate": 1.9149511704925945e-06,
1779
+ "loss": 0.0436,
1780
+ "step": 253
1781
+ },
1782
+ {
1783
+ "epoch": 4.0978260869565215,
1784
+ "grad_norm": 0.24933360517024994,
1785
+ "learning_rate": 1.8862862821480023e-06,
1786
+ "loss": 0.0409,
1787
+ "step": 254
1788
+ },
1789
+ {
1790
+ "epoch": 4.114130434782608,
1791
+ "grad_norm": 0.24037982523441315,
1792
+ "learning_rate": 1.8577069984334522e-06,
1793
+ "loss": 0.0445,
1794
+ "step": 255
1795
+ },
1796
+ {
1797
+ "epoch": 4.130434782608695,
1798
+ "grad_norm": 0.22139646112918854,
1799
+ "learning_rate": 1.829217305766289e-06,
1800
+ "loss": 0.041,
1801
+ "step": 256
1802
+ },
1803
+ {
1804
+ "epoch": 4.146739130434782,
1805
+ "grad_norm": 0.23465107381343842,
1806
+ "learning_rate": 1.8008211780671353e-06,
1807
+ "loss": 0.0456,
1808
+ "step": 257
1809
+ },
1810
+ {
1811
+ "epoch": 4.163043478260869,
1812
+ "grad_norm": 0.23133054375648499,
1813
+ "learning_rate": 1.772522576205589e-06,
1814
+ "loss": 0.0428,
1815
+ "step": 258
1816
+ },
1817
+ {
1818
+ "epoch": 4.179347826086956,
1819
+ "grad_norm": 0.25171801447868347,
1820
+ "learning_rate": 1.7443254474477328e-06,
1821
+ "loss": 0.0387,
1822
+ "step": 259
1823
+ },
1824
+ {
1825
+ "epoch": 4.195652173913044,
1826
+ "grad_norm": 0.22239144146442413,
1827
+ "learning_rate": 1.7162337249055478e-06,
1828
+ "loss": 0.0417,
1829
+ "step": 260
1830
+ },
1831
+ {
1832
+ "epoch": 4.211956521739131,
1833
+ "grad_norm": 0.24063155055046082,
1834
+ "learning_rate": 1.6882513269882916e-06,
1835
+ "loss": 0.0451,
1836
+ "step": 261
1837
+ },
1838
+ {
1839
+ "epoch": 4.228260869565218,
1840
+ "grad_norm": 0.24732705950737,
1841
+ "learning_rate": 1.6603821568559436e-06,
1842
+ "loss": 0.0428,
1843
+ "step": 262
1844
+ },
1845
+ {
1846
+ "epoch": 4.244565217391305,
1847
+ "grad_norm": 0.2546471953392029,
1848
+ "learning_rate": 1.6326301018747623e-06,
1849
+ "loss": 0.0378,
1850
+ "step": 263
1851
+ },
1852
+ {
1853
+ "epoch": 4.260869565217392,
1854
+ "grad_norm": 0.2471722811460495,
1855
+ "learning_rate": 1.6049990330750508e-06,
1856
+ "loss": 0.0401,
1857
+ "step": 264
1858
+ },
1859
+ {
1860
+ "epoch": 4.2771739130434785,
1861
+ "grad_norm": 0.20250661671161652,
1862
+ "learning_rate": 1.5774928046112025e-06,
1863
+ "loss": 0.0404,
1864
+ "step": 265
1865
+ },
1866
+ {
1867
+ "epoch": 4.293478260869565,
1868
+ "grad_norm": 0.21679657697677612,
1869
+ "learning_rate": 1.5501152532241003e-06,
1870
+ "loss": 0.0432,
1871
+ "step": 266
1872
+ },
1873
+ {
1874
+ "epoch": 4.309782608695652,
1875
+ "grad_norm": 0.25477948784828186,
1876
+ "learning_rate": 1.522870197705943e-06,
1877
+ "loss": 0.0484,
1878
+ "step": 267
1879
+ },
1880
+ {
1881
+ "epoch": 4.326086956521739,
1882
+ "grad_norm": 0.23823678493499756,
1883
+ "learning_rate": 1.495761438367577e-06,
1884
+ "loss": 0.038,
1885
+ "step": 268
1886
+ },
1887
+ {
1888
+ "epoch": 4.342391304347826,
1889
+ "grad_norm": 0.19264428317546844,
1890
+ "learning_rate": 1.4687927565084023e-06,
1891
+ "loss": 0.0463,
1892
+ "step": 269
1893
+ },
1894
+ {
1895
+ "epoch": 4.358695652173913,
1896
+ "grad_norm": 0.2232806533575058,
1897
+ "learning_rate": 1.4419679138889379e-06,
1898
+ "loss": 0.0367,
1899
+ "step": 270
1900
+ },
1901
+ {
1902
+ "epoch": 4.375,
1903
+ "grad_norm": 0.227358877658844,
1904
+ "learning_rate": 1.415290652206105e-06,
1905
+ "loss": 0.0393,
1906
+ "step": 271
1907
+ },
1908
+ {
1909
+ "epoch": 4.391304347826087,
1910
+ "grad_norm": 0.21266399323940277,
1911
+ "learning_rate": 1.3887646925713116e-06,
1912
+ "loss": 0.0427,
1913
+ "step": 272
1914
+ },
1915
+ {
1916
+ "epoch": 4.407608695652174,
1917
+ "grad_norm": 0.219986230134964,
1918
+ "learning_rate": 1.3623937349914093e-06,
1919
+ "loss": 0.0391,
1920
+ "step": 273
1921
+ },
1922
+ {
1923
+ "epoch": 4.423913043478261,
1924
+ "grad_norm": 0.20911090075969696,
1925
+ "learning_rate": 1.3361814578525922e-06,
1926
+ "loss": 0.0414,
1927
+ "step": 274
1928
+ },
1929
+ {
1930
+ "epoch": 4.440217391304348,
1931
+ "grad_norm": 0.20766796171665192,
1932
+ "learning_rate": 1.3101315174073162e-06,
1933
+ "loss": 0.0427,
1934
+ "step": 275
1935
+ },
1936
+ {
1937
+ "epoch": 4.456521739130435,
1938
+ "grad_norm": 0.2213449478149414,
1939
+ "learning_rate": 1.2842475472642969e-06,
1940
+ "loss": 0.0412,
1941
+ "step": 276
1942
+ },
1943
+ {
1944
+ "epoch": 4.4728260869565215,
1945
+ "grad_norm": 0.21896588802337646,
1946
+ "learning_rate": 1.258533157881674e-06,
1947
+ "loss": 0.0419,
1948
+ "step": 277
1949
+ },
1950
+ {
1951
+ "epoch": 4.489130434782608,
1952
+ "grad_norm": 0.23026618361473083,
1953
+ "learning_rate": 1.2329919360634003e-06,
1954
+ "loss": 0.0391,
1955
+ "step": 278
1956
+ },
1957
+ {
1958
+ "epoch": 4.505434782608695,
1959
+ "grad_norm": 0.2366199791431427,
1960
+ "learning_rate": 1.2076274444589361e-06,
1961
+ "loss": 0.0451,
1962
+ "step": 279
1963
+ },
1964
+ {
1965
+ "epoch": 4.521739130434782,
1966
+ "grad_norm": 0.21407200396060944,
1967
+ "learning_rate": 1.182443221066303e-06,
1968
+ "loss": 0.0445,
1969
+ "step": 280
1970
+ },
1971
+ {
1972
+ "epoch": 4.538043478260869,
1973
+ "grad_norm": 0.22052663564682007,
1974
+ "learning_rate": 1.1574427787385853e-06,
1975
+ "loss": 0.0454,
1976
+ "step": 281
1977
+ },
1978
+ {
1979
+ "epoch": 4.554347826086957,
1980
+ "grad_norm": 0.23922421038150787,
1981
+ "learning_rate": 1.1326296046939334e-06,
1982
+ "loss": 0.0386,
1983
+ "step": 282
1984
+ },
1985
+ {
1986
+ "epoch": 4.570652173913043,
1987
+ "grad_norm": 0.21938639879226685,
1988
+ "learning_rate": 1.1080071600291453e-06,
1989
+ "loss": 0.0424,
1990
+ "step": 283
1991
+ },
1992
+ {
1993
+ "epoch": 4.586956521739131,
1994
+ "grad_norm": 0.23158493638038635,
1995
+ "learning_rate": 1.083578879236895e-06,
1996
+ "loss": 0.0425,
1997
+ "step": 284
1998
+ },
1999
+ {
2000
+ "epoch": 4.603260869565218,
2001
+ "grad_norm": 0.22377173602581024,
2002
+ "learning_rate": 1.0593481697266582e-06,
2003
+ "loss": 0.0431,
2004
+ "step": 285
2005
+ },
2006
+ {
2007
+ "epoch": 4.619565217391305,
2008
+ "grad_norm": 0.21283438801765442,
2009
+ "learning_rate": 1.0353184113494386e-06,
2010
+ "loss": 0.0383,
2011
+ "step": 286
2012
+ },
2013
+ {
2014
+ "epoch": 4.635869565217392,
2015
+ "grad_norm": 0.24626271426677704,
2016
+ "learning_rate": 1.0114929559263122e-06,
2017
+ "loss": 0.04,
2018
+ "step": 287
2019
+ },
2020
+ {
2021
+ "epoch": 4.6521739130434785,
2022
+ "grad_norm": 0.23055487871170044,
2023
+ "learning_rate": 9.878751267809069e-07,
2024
+ "loss": 0.0438,
2025
+ "step": 288
2026
+ },
2027
+ {
2028
+ "epoch": 4.668478260869565,
2029
+ "grad_norm": 0.21289804577827454,
2030
+ "learning_rate": 9.644682182758305e-07,
2031
+ "loss": 0.044,
2032
+ "step": 289
2033
+ },
2034
+ {
2035
+ "epoch": 4.684782608695652,
2036
+ "grad_norm": 0.25633224844932556,
2037
+ "learning_rate": 9.412754953531664e-07,
2038
+ "loss": 0.0387,
2039
+ "step": 290
2040
+ },
2041
+ {
2042
+ "epoch": 4.701086956521739,
2043
+ "grad_norm": 0.21574127674102783,
2044
+ "learning_rate": 9.183001930790483e-07,
2045
+ "loss": 0.0408,
2046
+ "step": 291
2047
+ },
2048
+ {
2049
+ "epoch": 4.717391304347826,
2050
+ "grad_norm": 0.2526251971721649,
2051
+ "learning_rate": 8.955455161924217e-07,
2052
+ "loss": 0.0431,
2053
+ "step": 292
2054
+ },
2055
+ {
2056
+ "epoch": 4.733695652173913,
2057
+ "grad_norm": 0.2219126969575882,
2058
+ "learning_rate": 8.730146386580157e-07,
2059
+ "loss": 0.0382,
2060
+ "step": 293
2061
+ },
2062
+ {
2063
+ "epoch": 4.75,
2064
+ "grad_norm": 0.22477059066295624,
2065
+ "learning_rate": 8.507107032236323e-07,
2066
+ "loss": 0.0387,
2067
+ "step": 294
2068
+ },
2069
+ {
2070
+ "epoch": 4.766304347826087,
2071
+ "grad_norm": 0.23549960553646088,
2072
+ "learning_rate": 8.286368209817644e-07,
2073
+ "loss": 0.042,
2074
+ "step": 295
2075
+ },
2076
+ {
2077
+ "epoch": 4.782608695652174,
2078
+ "grad_norm": 0.24367327988147736,
2079
+ "learning_rate": 8.067960709356479e-07,
2080
+ "loss": 0.0405,
2081
+ "step": 296
2082
+ },
2083
+ {
2084
+ "epoch": 4.798913043478261,
2085
+ "grad_norm": 0.25587689876556396,
2086
+ "learning_rate": 7.851914995697801e-07,
2087
+ "loss": 0.0416,
2088
+ "step": 297
2089
+ },
2090
+ {
2091
+ "epoch": 4.815217391304348,
2092
+ "grad_norm": 0.270684152841568,
2093
+ "learning_rate": 7.638261204249783e-07,
2094
+ "loss": 0.0394,
2095
+ "step": 298
2096
+ },
2097
+ {
2098
+ "epoch": 4.831521739130435,
2099
+ "grad_norm": 0.22876906394958496,
2100
+ "learning_rate": 7.427029136780333e-07,
2101
+ "loss": 0.0369,
2102
+ "step": 299
2103
+ },
2104
+ {
2105
+ "epoch": 4.8478260869565215,
2106
+ "grad_norm": 0.29206568002700806,
2107
+ "learning_rate": 7.218248257260127e-07,
2108
+ "loss": 0.0503,
2109
+ "step": 300
2110
+ },
2111
+ {
2112
+ "epoch": 4.864130434782608,
2113
+ "grad_norm": 0.22232329845428467,
2114
+ "learning_rate": 7.011947687752804e-07,
2115
+ "loss": 0.0407,
2116
+ "step": 301
2117
+ },
2118
+ {
2119
+ "epoch": 4.880434782608695,
2120
+ "grad_norm": 0.2493494153022766,
2121
+ "learning_rate": 6.808156204352845e-07,
2122
+ "loss": 0.0394,
2123
+ "step": 302
2124
+ },
2125
+ {
2126
+ "epoch": 4.896739130434782,
2127
+ "grad_norm": 0.20831190049648285,
2128
+ "learning_rate": 6.60690223317171e-07,
2129
+ "loss": 0.0406,
2130
+ "step": 303
2131
+ },
2132
+ {
2133
+ "epoch": 4.913043478260869,
2134
+ "grad_norm": 0.22676558792591095,
2135
+ "learning_rate": 6.40821384637276e-07,
2136
+ "loss": 0.0365,
2137
+ "step": 304
2138
+ },
2139
+ {
2140
+ "epoch": 4.929347826086957,
2141
+ "grad_norm": 0.2518041133880615,
2142
+ "learning_rate": 6.212118758255595e-07,
2143
+ "loss": 0.0475,
2144
+ "step": 305
2145
+ },
2146
+ {
2147
+ "epoch": 4.945652173913043,
2148
+ "grad_norm": 0.23944813013076782,
2149
+ "learning_rate": 6.018644321390288e-07,
2150
+ "loss": 0.042,
2151
+ "step": 306
2152
+ },
2153
+ {
2154
+ "epoch": 4.961956521739131,
2155
+ "grad_norm": 0.2410481721162796,
2156
+ "learning_rate": 5.827817522802065e-07,
2157
+ "loss": 0.0389,
2158
+ "step": 307
2159
+ },
2160
+ {
2161
+ "epoch": 4.978260869565218,
2162
+ "grad_norm": 0.2525772750377655,
2163
+ "learning_rate": 5.639664980207024e-07,
2164
+ "loss": 0.0406,
2165
+ "step": 308
2166
+ },
2167
+ {
2168
+ "epoch": 4.994565217391305,
2169
+ "grad_norm": 0.276832640171051,
2170
+ "learning_rate": 5.454212938299256e-07,
2171
+ "loss": 0.0439,
2172
+ "step": 309
2173
+ },
2174
+ {
2175
+ "epoch": 5.0,
2176
+ "grad_norm": 0.276832640171051,
2177
+ "learning_rate": 5.271487265090163e-07,
2178
+ "loss": 0.0354,
2179
+ "step": 310
2180
+ },
2181
+ {
2182
+ "epoch": 5.016304347826087,
2183
+ "grad_norm": 0.5199080109596252,
2184
+ "learning_rate": 5.091513448300142e-07,
2185
+ "loss": 0.0407,
2186
+ "step": 311
2187
+ },
2188
+ {
2189
+ "epoch": 5.032608695652174,
2190
+ "grad_norm": 0.270175963640213,
2191
+ "learning_rate": 4.914316591803475e-07,
2192
+ "loss": 0.0402,
2193
+ "step": 312
2194
+ },
2195
+ {
2196
+ "epoch": 5.048913043478261,
2197
+ "grad_norm": 0.22836650907993317,
2198
+ "learning_rate": 4.739921412126591e-07,
2199
+ "loss": 0.0382,
2200
+ "step": 313
2201
+ },
2202
+ {
2203
+ "epoch": 5.065217391304348,
2204
+ "grad_norm": 0.22557897865772247,
2205
+ "learning_rate": 4.5683522350005505e-07,
2206
+ "loss": 0.0388,
2207
+ "step": 314
2208
+ },
2209
+ {
2210
+ "epoch": 5.081521739130435,
2211
+ "grad_norm": 0.23105387389659882,
2212
+ "learning_rate": 4.399632991967867e-07,
2213
+ "loss": 0.0362,
2214
+ "step": 315
2215
+ },
2216
+ {
2217
+ "epoch": 5.0978260869565215,
2218
+ "grad_norm": 0.26361796259880066,
2219
+ "learning_rate": 4.23378721704443e-07,
2220
+ "loss": 0.0466,
2221
+ "step": 316
2222
+ },
2223
+ {
2224
+ "epoch": 5.114130434782608,
2225
+ "grad_norm": 0.2395378202199936,
2226
+ "learning_rate": 4.070838043436787e-07,
2227
+ "loss": 0.041,
2228
+ "step": 317
2229
+ },
2230
+ {
2231
+ "epoch": 5.130434782608695,
2232
+ "grad_norm": 0.20131447911262512,
2233
+ "learning_rate": 3.910808200315433e-07,
2234
+ "loss": 0.0388,
2235
+ "step": 318
2236
+ },
2237
+ {
2238
+ "epoch": 5.146739130434782,
2239
+ "grad_norm": 0.2295176386833191,
2240
+ "learning_rate": 3.753720009644371e-07,
2241
+ "loss": 0.037,
2242
+ "step": 319
2243
+ },
2244
+ {
2245
+ "epoch": 5.163043478260869,
2246
+ "grad_norm": 0.24505485594272614,
2247
+ "learning_rate": 3.5995953830675004e-07,
2248
+ "loss": 0.043,
2249
+ "step": 320
2250
+ },
2251
+ {
2252
+ "epoch": 5.179347826086956,
2253
+ "grad_norm": 0.2312752902507782,
2254
+ "learning_rate": 3.448455818852267e-07,
2255
+ "loss": 0.0369,
2256
+ "step": 321
2257
+ },
2258
+ {
2259
+ "epoch": 5.195652173913044,
2260
+ "grad_norm": 0.21873357892036438,
2261
+ "learning_rate": 3.3003223988909234e-07,
2262
+ "loss": 0.0398,
2263
+ "step": 322
2264
+ },
2265
+ {
2266
+ "epoch": 5.211956521739131,
2267
+ "grad_norm": 0.22271527349948883,
2268
+ "learning_rate": 3.1552157857599327e-07,
2269
+ "loss": 0.0377,
2270
+ "step": 323
2271
+ },
2272
+ {
2273
+ "epoch": 5.228260869565218,
2274
+ "grad_norm": 0.22470292448997498,
2275
+ "learning_rate": 3.0131562198377763e-07,
2276
+ "loss": 0.0411,
2277
+ "step": 324
2278
+ },
2279
+ {
2280
+ "epoch": 5.244565217391305,
2281
+ "grad_norm": 0.22748963534832,
2282
+ "learning_rate": 2.874163516481732e-07,
2283
+ "loss": 0.0346,
2284
+ "step": 325
2285
+ },
2286
+ {
2287
+ "epoch": 5.260869565217392,
2288
+ "grad_norm": 0.27392780780792236,
2289
+ "learning_rate": 2.7382570632638853e-07,
2290
+ "loss": 0.0426,
2291
+ "step": 326
2292
+ },
2293
+ {
2294
+ "epoch": 5.2771739130434785,
2295
+ "grad_norm": 0.2509923577308655,
2296
+ "learning_rate": 2.605455817266861e-07,
2297
+ "loss": 0.0375,
2298
+ "step": 327
2299
+ },
2300
+ {
2301
+ "epoch": 5.293478260869565,
2302
+ "grad_norm": 0.20978674292564392,
2303
+ "learning_rate": 2.4757783024395244e-07,
2304
+ "loss": 0.0333,
2305
+ "step": 328
2306
+ },
2307
+ {
2308
+ "epoch": 5.309782608695652,
2309
+ "grad_norm": 0.23130132257938385,
2310
+ "learning_rate": 2.3492426070131746e-07,
2311
+ "loss": 0.0429,
2312
+ "step": 329
2313
+ },
2314
+ {
2315
+ "epoch": 5.326086956521739,
2316
+ "grad_norm": 0.2261745035648346,
2317
+ "learning_rate": 2.2258663809784892e-07,
2318
+ "loss": 0.0376,
2319
+ "step": 330
2320
+ },
2321
+ {
2322
+ "epoch": 5.342391304347826,
2323
+ "grad_norm": 0.21033048629760742,
2324
+ "learning_rate": 2.1056668336235624e-07,
2325
+ "loss": 0.039,
2326
+ "step": 331
2327
+ },
2328
+ {
2329
+ "epoch": 5.358695652173913,
2330
+ "grad_norm": 0.23950928449630737,
2331
+ "learning_rate": 1.9886607311334987e-07,
2332
+ "loss": 0.0375,
2333
+ "step": 332
2334
+ },
2335
+ {
2336
+ "epoch": 5.375,
2337
+ "grad_norm": 0.2400600165128708,
2338
+ "learning_rate": 1.8748643942516882e-07,
2339
+ "loss": 0.0447,
2340
+ "step": 333
2341
+ },
2342
+ {
2343
+ "epoch": 5.391304347826087,
2344
+ "grad_norm": 0.24937079846858978,
2345
+ "learning_rate": 1.764293696003358e-07,
2346
+ "loss": 0.0439,
2347
+ "step": 334
2348
+ },
2349
+ {
2350
+ "epoch": 5.407608695652174,
2351
+ "grad_norm": 0.2523324489593506,
2352
+ "learning_rate": 1.656964059481453e-07,
2353
+ "loss": 0.0368,
2354
+ "step": 335
2355
+ },
2356
+ {
2357
+ "epoch": 5.423913043478261,
2358
+ "grad_norm": 0.24901482462882996,
2359
+ "learning_rate": 1.552890455695369e-07,
2360
+ "loss": 0.0382,
2361
+ "step": 336
2362
+ },
2363
+ {
2364
+ "epoch": 5.440217391304348,
2365
+ "grad_norm": 0.21542105078697205,
2366
+ "learning_rate": 1.4520874014826464e-07,
2367
+ "loss": 0.04,
2368
+ "step": 337
2369
+ },
2370
+ {
2371
+ "epoch": 5.456521739130435,
2372
+ "grad_norm": 0.24464815855026245,
2373
+ "learning_rate": 1.3545689574841341e-07,
2374
+ "loss": 0.0358,
2375
+ "step": 338
2376
+ },
2377
+ {
2378
+ "epoch": 5.4728260869565215,
2379
+ "grad_norm": 0.24528922140598297,
2380
+ "learning_rate": 1.2603487261826726e-07,
2381
+ "loss": 0.038,
2382
+ "step": 339
2383
+ },
2384
+ {
2385
+ "epoch": 5.489130434782608,
2386
+ "grad_norm": 0.22918996214866638,
2387
+ "learning_rate": 1.1694398500057714e-07,
2388
+ "loss": 0.0421,
2389
+ "step": 340
2390
+ },
2391
+ {
2392
+ "epoch": 5.505434782608695,
2393
+ "grad_norm": 0.2410709410905838,
2394
+ "learning_rate": 1.081855009492383e-07,
2395
+ "loss": 0.0395,
2396
+ "step": 341
2397
+ },
2398
+ {
2399
+ "epoch": 5.521739130434782,
2400
+ "grad_norm": 0.22620242834091187,
2401
+ "learning_rate": 9.976064215241859e-08,
2402
+ "loss": 0.0392,
2403
+ "step": 342
2404
+ },
2405
+ {
2406
+ "epoch": 5.538043478260869,
2407
+ "grad_norm": 0.24372537434101105,
2408
+ "learning_rate": 9.167058376214621e-08,
2409
+ "loss": 0.0379,
2410
+ "step": 343
2411
+ },
2412
+ {
2413
+ "epoch": 5.554347826086957,
2414
+ "grad_norm": 0.21486566960811615,
2415
+ "learning_rate": 8.391645423039357e-08,
2416
+ "loss": 0.0377,
2417
+ "step": 344
2418
+ },
2419
+ {
2420
+ "epoch": 5.570652173913043,
2421
+ "grad_norm": 0.22310517728328705,
2422
+ "learning_rate": 7.649933515167407e-08,
2423
+ "loss": 0.0386,
2424
+ "step": 345
2425
+ },
2426
+ {
2427
+ "epoch": 5.586956521739131,
2428
+ "grad_norm": 0.23468376696109772,
2429
+ "learning_rate": 6.94202611121736e-08,
2430
+ "loss": 0.0398,
2431
+ "step": 346
2432
+ },
2433
+ {
2434
+ "epoch": 5.603260869565218,
2435
+ "grad_norm": 0.24592317640781403,
2436
+ "learning_rate": 6.268021954544095e-08,
2437
+ "loss": 0.0346,
2438
+ "step": 347
2439
+ },
2440
+ {
2441
+ "epoch": 5.619565217391305,
2442
+ "grad_norm": 0.2240305244922638,
2443
+ "learning_rate": 5.628015059465364e-08,
2444
+ "loss": 0.0483,
2445
+ "step": 348
2446
+ },
2447
+ {
2448
+ "epoch": 5.635869565217392,
2449
+ "grad_norm": 0.2843106687068939,
2450
+ "learning_rate": 5.022094698148072e-08,
2451
+ "loss": 0.0427,
2452
+ "step": 349
2453
+ },
2454
+ {
2455
+ "epoch": 5.6521739130434785,
2456
+ "grad_norm": 0.2484791874885559,
2457
+ "learning_rate": 4.450345388156141e-08,
2458
+ "loss": 0.0429,
2459
+ "step": 350
2460
+ },
2461
+ {
2462
+ "epoch": 5.668478260869565,
2463
+ "grad_norm": 0.23303182423114777,
2464
+ "learning_rate": 3.9128468806614304e-08,
2465
+ "loss": 0.0355,
2466
+ "step": 351
2467
+ },
2468
+ {
2469
+ "epoch": 5.684782608695652,
2470
+ "grad_norm": 0.22100433707237244,
2471
+ "learning_rate": 3.4096741493194196e-08,
2472
+ "loss": 0.0386,
2473
+ "step": 352
2474
+ },
2475
+ {
2476
+ "epoch": 5.701086956521739,
2477
+ "grad_norm": 0.2393917292356491,
2478
+ "learning_rate": 2.940897379811597e-08,
2479
+ "loss": 0.0495,
2480
+ "step": 353
2481
+ },
2482
+ {
2483
+ "epoch": 5.717391304347826,
2484
+ "grad_norm": 0.27596771717071533,
2485
+ "learning_rate": 2.506581960055432e-08,
2486
+ "loss": 0.0388,
2487
+ "step": 354
2488
+ },
2489
+ {
2490
+ "epoch": 5.733695652173913,
2491
+ "grad_norm": 0.23731505870819092,
2492
+ "learning_rate": 2.106788471083615e-08,
2493
+ "loss": 0.039,
2494
+ "step": 355
2495
+ },
2496
+ {
2497
+ "epoch": 5.75,
2498
+ "grad_norm": 0.23420853912830353,
2499
+ "learning_rate": 1.7415726785939836e-08,
2500
+ "loss": 0.0466,
2501
+ "step": 356
2502
+ },
2503
+ {
2504
+ "epoch": 5.766304347826087,
2505
+ "grad_norm": 0.2571556866168976,
2506
+ "learning_rate": 1.4109855251708272e-08,
2507
+ "loss": 0.0398,
2508
+ "step": 357
2509
+ },
2510
+ {
2511
+ "epoch": 5.782608695652174,
2512
+ "grad_norm": 0.25577929615974426,
2513
+ "learning_rate": 1.115073123179128e-08,
2514
+ "loss": 0.0418,
2515
+ "step": 358
2516
+ },
2517
+ {
2518
+ "epoch": 5.798913043478261,
2519
+ "grad_norm": 0.25703752040863037,
2520
+ "learning_rate": 8.538767483325384e-09,
2521
+ "loss": 0.0382,
2522
+ "step": 359
2523
+ },
2524
+ {
2525
+ "epoch": 5.815217391304348,
2526
+ "grad_norm": 0.21773234009742737,
2527
+ "learning_rate": 6.274328339360702e-09,
2528
+ "loss": 0.0374,
2529
+ "step": 360
2530
+ },
2531
+ {
2532
+ "epoch": 5.831521739130435,
2533
+ "grad_norm": 0.24684062600135803,
2534
+ "learning_rate": 4.357729658039378e-09,
2535
+ "loss": 0.041,
2536
+ "step": 361
2537
+ },
2538
+ {
2539
+ "epoch": 5.8478260869565215,
2540
+ "grad_norm": 0.20334966480731964,
2541
+ "learning_rate": 2.789238778540537e-09,
2542
+ "loss": 0.0386,
2543
+ "step": 362
2544
+ },
2545
+ {
2546
+ "epoch": 5.864130434782608,
2547
+ "grad_norm": 0.2612401247024536,
2548
+ "learning_rate": 1.5690744837873473e-09,
2549
+ "loss": 0.0415,
2550
+ "step": 363
2551
+ },
2552
+ {
2553
+ "epoch": 5.880434782608695,
2554
+ "grad_norm": 0.2394614964723587,
2555
+ "learning_rate": 6.974069699314246e-10,
2556
+ "loss": 0.0376,
2557
+ "step": 364
2558
+ },
2559
+ {
2560
+ "epoch": 5.896739130434782,
2561
+ "grad_norm": 0.23148567974567413,
2562
+ "learning_rate": 1.743578226129361e-10,
2563
+ "loss": 0.0387,
2564
+ "step": 365
2565
+ },
2566
+ {
2567
+ "epoch": 5.913043478260869,
2568
+ "grad_norm": 0.24650397896766663,
2569
+ "learning_rate": 0.0,
2570
+ "loss": 0.0396,
2571
+ "step": 366
2572
+ }
2573
+ ],
2574
+ "logging_steps": 1,
2575
+ "max_steps": 366,
2576
+ "num_input_tokens_seen": 0,
2577
+ "num_train_epochs": 6,
2578
+ "save_steps": 61,
2579
+ "stateful_callbacks": {
2580
+ "TrainerControl": {
2581
+ "args": {
2582
+ "should_epoch_stop": false,
2583
+ "should_evaluate": false,
2584
+ "should_log": false,
2585
+ "should_save": true,
2586
+ "should_training_stop": true
2587
+ },
2588
+ "attributes": {}
2589
+ }
2590
+ },
2591
+ "total_flos": 8.114918032020603e+17,
2592
+ "train_batch_size": 4,
2593
+ "trial_name": null,
2594
+ "trial_params": null
2595
+ }
checkpoint-366/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4bdf3f34ad3174d7ffb541641f906b4ea87bd3f0ed6dc8fb586e66791457854
3
+ size 7992
checkpoint-366/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-366/zero_to_fp32.py ADDED
@@ -0,0 +1,604 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
215
+ exclude_frozen_parameters)
216
+ elif zero_stage == 3:
217
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
218
+ exclude_frozen_parameters)
219
+
220
+
221
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
222
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
223
+ return
224
+
225
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
226
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
227
+
228
+ if debug:
229
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
230
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
231
+
232
+ wanted_params = len(frozen_param_shapes)
233
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
234
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
235
+ print(f'Frozen params: Have {avail_numel} numels to process.')
236
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
237
+
238
+ total_params = 0
239
+ total_numel = 0
240
+ for name, shape in frozen_param_shapes.items():
241
+ total_params += 1
242
+ unpartitioned_numel = shape.numel()
243
+ total_numel += unpartitioned_numel
244
+
245
+ state_dict[name] = frozen_param_fragments[name]
246
+
247
+ if debug:
248
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
249
+
250
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
251
+
252
+
253
+ def _has_callable(obj, fn):
254
+ attr = getattr(obj, fn, None)
255
+ return callable(attr)
256
+
257
+
258
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
259
+ param_shapes = zero_model_states[0].param_shapes
260
+
261
+ # Reconstruction protocol:
262
+ #
263
+ # XXX: document this
264
+
265
+ if debug:
266
+ for i in range(world_size):
267
+ for j in range(len(fp32_flat_groups[0])):
268
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
269
+
270
+ # XXX: memory usage doubles here (zero2)
271
+ num_param_groups = len(fp32_flat_groups[0])
272
+ merged_single_partition_of_fp32_groups = []
273
+ for i in range(num_param_groups):
274
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
275
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
276
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
277
+ avail_numel = sum(
278
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
279
+
280
+ if debug:
281
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
282
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
283
+ # not asserting if there is a mismatch due to possible padding
284
+ print(f"Have {avail_numel} numels to process.")
285
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
286
+
287
+ # params
288
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
289
+ # out-of-core computing solution
290
+ total_numel = 0
291
+ total_params = 0
292
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
293
+ offset = 0
294
+ avail_numel = full_single_fp32_vector.numel()
295
+ for name, shape in shapes.items():
296
+
297
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
298
+ total_numel += unpartitioned_numel
299
+ total_params += 1
300
+
301
+ if debug:
302
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
303
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
304
+ offset += unpartitioned_numel
305
+
306
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
307
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
308
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
309
+ # live optimizer object, so we are checking that the numbers are within the right range
310
+ align_to = 2 * world_size
311
+
312
+ def zero2_align(x):
313
+ return align_to * math.ceil(x / align_to)
314
+
315
+ if debug:
316
+ print(f"original offset={offset}, avail_numel={avail_numel}")
317
+
318
+ offset = zero2_align(offset)
319
+ avail_numel = zero2_align(avail_numel)
320
+
321
+ if debug:
322
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
323
+
324
+ # Sanity check
325
+ if offset != avail_numel:
326
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
327
+
328
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
329
+
330
+
331
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
332
+ exclude_frozen_parameters):
333
+ state_dict = OrderedDict()
334
+
335
+ # buffers
336
+ buffers = zero_model_states[0].buffers
337
+ state_dict.update(buffers)
338
+ if debug:
339
+ print(f"added {len(buffers)} buffers")
340
+
341
+ if not exclude_frozen_parameters:
342
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
343
+
344
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
345
+
346
+ # recover shared parameters
347
+ for pair in zero_model_states[0].shared_params:
348
+ if pair[1] in state_dict:
349
+ state_dict[pair[0]] = state_dict[pair[1]]
350
+
351
+ return state_dict
352
+
353
+
354
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
355
+ remainder = unpartitioned_numel % world_size
356
+ padding_numel = (world_size - remainder) if remainder else 0
357
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
358
+ return partitioned_numel, padding_numel
359
+
360
+
361
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
362
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
363
+ return
364
+
365
+ if debug:
366
+ for i in range(world_size):
367
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
368
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
369
+
370
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
371
+ wanted_params = len(frozen_param_shapes)
372
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
373
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
374
+ print(f'Frozen params: Have {avail_numel} numels to process.')
375
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
376
+
377
+ total_params = 0
378
+ total_numel = 0
379
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
380
+ total_params += 1
381
+ unpartitioned_numel = shape.numel()
382
+ total_numel += unpartitioned_numel
383
+
384
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
385
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
386
+
387
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
388
+
389
+ if debug:
390
+ print(
391
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
392
+ )
393
+
394
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
395
+
396
+
397
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
398
+ param_shapes = zero_model_states[0].param_shapes
399
+ avail_numel = fp32_flat_groups[0].numel() * world_size
400
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
401
+ # param, re-consolidating each param, while dealing with padding if any
402
+
403
+ # merge list of dicts, preserving order
404
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
405
+
406
+ if debug:
407
+ for i in range(world_size):
408
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
409
+
410
+ wanted_params = len(param_shapes)
411
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
412
+ # not asserting if there is a mismatch due to possible padding
413
+ avail_numel = fp32_flat_groups[0].numel() * world_size
414
+ print(f"Trainable params: Have {avail_numel} numels to process.")
415
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
416
+
417
+ # params
418
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
419
+ # out-of-core computing solution
420
+ offset = 0
421
+ total_numel = 0
422
+ total_params = 0
423
+ for name, shape in param_shapes.items():
424
+
425
+ unpartitioned_numel = shape.numel()
426
+ total_numel += unpartitioned_numel
427
+ total_params += 1
428
+
429
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
430
+
431
+ if debug:
432
+ print(
433
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
434
+ )
435
+
436
+ # XXX: memory usage doubles here
437
+ state_dict[name] = torch.cat(
438
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
439
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
440
+ offset += partitioned_numel
441
+
442
+ offset *= world_size
443
+
444
+ # Sanity check
445
+ if offset != avail_numel:
446
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
447
+
448
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
449
+
450
+
451
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
452
+ exclude_frozen_parameters):
453
+ state_dict = OrderedDict()
454
+
455
+ # buffers
456
+ buffers = zero_model_states[0].buffers
457
+ state_dict.update(buffers)
458
+ if debug:
459
+ print(f"added {len(buffers)} buffers")
460
+
461
+ if not exclude_frozen_parameters:
462
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
463
+
464
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
465
+
466
+ # recover shared parameters
467
+ for pair in zero_model_states[0].shared_params:
468
+ if pair[1] in state_dict:
469
+ state_dict[pair[0]] = state_dict[pair[1]]
470
+
471
+ return state_dict
472
+
473
+
474
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
475
+ """
476
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
477
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
478
+ via a model hub.
479
+
480
+ Args:
481
+ - ``checkpoint_dir``: path to the desired checkpoint folder
482
+ - ``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``
483
+ - ``exclude_frozen_parameters``: exclude frozen parameters
484
+
485
+ Returns:
486
+ - pytorch ``state_dict``
487
+
488
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
489
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
490
+ the checkpoint.
491
+
492
+ A typical usage might be ::
493
+
494
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
495
+ # do the training and checkpoint saving
496
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
497
+ model = model.cpu() # move to cpu
498
+ model.load_state_dict(state_dict)
499
+ # submit to model hub or save the model to share with others
500
+
501
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
502
+ application. i.e. you will need to re-initialize the deepspeed engine, since
503
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
504
+
505
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
506
+
507
+ """
508
+ if tag is None:
509
+ latest_path = os.path.join(checkpoint_dir, 'latest')
510
+ if os.path.isfile(latest_path):
511
+ with open(latest_path, 'r') as fd:
512
+ tag = fd.read().strip()
513
+ else:
514
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
515
+
516
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
517
+
518
+ if not os.path.isdir(ds_checkpoint_dir):
519
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
520
+
521
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
522
+
523
+
524
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
525
+ """
526
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
527
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
528
+
529
+ Args:
530
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
531
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
532
+ - ``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``
533
+ - ``exclude_frozen_parameters``: exclude frozen parameters
534
+ """
535
+
536
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
537
+ print(f"Saving fp32 state dict to {output_file}")
538
+ torch.save(state_dict, output_file)
539
+
540
+
541
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
542
+ """
543
+ 1. Put the provided model to cpu
544
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
545
+ 3. Load it into the provided model
546
+
547
+ Args:
548
+ - ``model``: the model object to update
549
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
550
+ - ``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``
551
+
552
+ Returns:
553
+ - ``model`: modified model
554
+
555
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
556
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
557
+ conveniently placed for you in the checkpoint folder.
558
+
559
+ A typical usage might be ::
560
+
561
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
562
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
563
+ # submit to model hub or save the model to share with others
564
+
565
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
566
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
567
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
568
+
569
+ """
570
+ logger.info(f"Extracting fp32 weights")
571
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
572
+
573
+ logger.info(f"Overwriting model with fp32 weights")
574
+ model = model.cpu()
575
+ model.load_state_dict(state_dict, strict=False)
576
+
577
+ return model
578
+
579
+
580
+ if __name__ == "__main__":
581
+
582
+ parser = argparse.ArgumentParser()
583
+ parser.add_argument("checkpoint_dir",
584
+ type=str,
585
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
586
+ parser.add_argument(
587
+ "output_file",
588
+ type=str,
589
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
590
+ parser.add_argument("-t",
591
+ "--tag",
592
+ type=str,
593
+ default=None,
594
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
595
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
596
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
597
+ args = parser.parse_args()
598
+
599
+ debug = args.debug
600
+
601
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
602
+ args.output_file,
603
+ tag=args.tag,
604
+ exclude_frozen_parameters=args.exclude_frozen_parameters)