chen commited on
Upload folder using huggingface_hub
Browse files- README.md +202 -0
- adapter_config.json +31 -0
- adapter_model.safetensors +3 -0
- additional_config.json +1 -0
- args.json +450 -0
- latest +1 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- trainer_state.json +2504 -0
- training_args.bin +3 -0
- zero_to_fp32.py +760 -0
README.md
ADDED
|
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: meta-llama/Llama-3.2-11B-Vision-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 0.15.2
|
adapter_config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alpha_pattern": {},
|
| 3 |
+
"auto_mapping": null,
|
| 4 |
+
"base_model_name_or_path": "/data1/chenjiazun/LLM/Llama-3.2-11B-Vision-Instruct",
|
| 5 |
+
"bias": "none",
|
| 6 |
+
"corda_config": null,
|
| 7 |
+
"eva_config": null,
|
| 8 |
+
"exclude_modules": null,
|
| 9 |
+
"fan_in_fan_out": false,
|
| 10 |
+
"inference_mode": true,
|
| 11 |
+
"init_lora_weights": true,
|
| 12 |
+
"layer_replication": null,
|
| 13 |
+
"layers_pattern": null,
|
| 14 |
+
"layers_to_transform": null,
|
| 15 |
+
"loftq_config": {},
|
| 16 |
+
"lora_alpha": 32,
|
| 17 |
+
"lora_bias": false,
|
| 18 |
+
"lora_dropout": 0.05,
|
| 19 |
+
"megatron_config": null,
|
| 20 |
+
"megatron_core": "megatron.core",
|
| 21 |
+
"modules_to_save": [],
|
| 22 |
+
"peft_type": "LORA",
|
| 23 |
+
"r": 8,
|
| 24 |
+
"rank_pattern": {},
|
| 25 |
+
"revision": null,
|
| 26 |
+
"target_modules": "(?!^(multi_modal_projector))^(language_model|vision_model).*\\.(gate_proj|k_proj|v_proj|q_proj|down_proj|o_proj|up_proj|k_proj|q_proj|v_proj|fc1|o_proj|fc2)$",
|
| 27 |
+
"task_type": "CAUSAL_LM",
|
| 28 |
+
"trainable_token_indices": null,
|
| 29 |
+
"use_dora": false,
|
| 30 |
+
"use_rslora": false
|
| 31 |
+
}
|
adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:457976987c513b4854bb529cd22f4a3238ad37381061fdbb0f1f4616887aee7d
|
| 3 |
+
size 67330312
|
additional_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"lora_dtype": null, "lorap_lr_ratio": null, "lorap_emb_lr": 1e-06}
|
args.json
ADDED
|
@@ -0,0 +1,450 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "/data1/chenjiazun/LLM/Llama-3.2-11B-Vision-Instruct",
|
| 3 |
+
"model_type": "llama3_2_vision",
|
| 4 |
+
"model_revision": null,
|
| 5 |
+
"task_type": "causal_lm",
|
| 6 |
+
"torch_dtype": "bfloat16",
|
| 7 |
+
"attn_impl": null,
|
| 8 |
+
"num_labels": null,
|
| 9 |
+
"problem_type": null,
|
| 10 |
+
"rope_scaling": null,
|
| 11 |
+
"device_map": null,
|
| 12 |
+
"max_memory": {},
|
| 13 |
+
"local_repo_path": null,
|
| 14 |
+
"template": "llama3_2_vision",
|
| 15 |
+
"system": null,
|
| 16 |
+
"max_length": 8092,
|
| 17 |
+
"truncation_strategy": "left",
|
| 18 |
+
"max_pixels": null,
|
| 19 |
+
"agent_template": null,
|
| 20 |
+
"norm_bbox": null,
|
| 21 |
+
"response_prefix": null,
|
| 22 |
+
"padding_side": "right",
|
| 23 |
+
"loss_scale": "last_round",
|
| 24 |
+
"sequence_parallel_size": 1,
|
| 25 |
+
"use_chat_template": true,
|
| 26 |
+
"template_backend": "swift",
|
| 27 |
+
"dataset": [
|
| 28 |
+
"./new_task1_nothink_respone_onlycorrect.json"
|
| 29 |
+
],
|
| 30 |
+
"val_dataset": [],
|
| 31 |
+
"split_dataset_ratio": 0.01,
|
| 32 |
+
"data_seed": 123,
|
| 33 |
+
"dataset_num_proc": 1,
|
| 34 |
+
"dataset_shuffle": true,
|
| 35 |
+
"val_dataset_shuffle": false,
|
| 36 |
+
"streaming": false,
|
| 37 |
+
"interleave_prob": null,
|
| 38 |
+
"stopping_strategy": "first_exhausted",
|
| 39 |
+
"shuffle_buffer_size": 1000,
|
| 40 |
+
"enable_cache": false,
|
| 41 |
+
"download_mode": "reuse_dataset_if_exists",
|
| 42 |
+
"columns": {},
|
| 43 |
+
"strict": false,
|
| 44 |
+
"remove_unused_columns": false,
|
| 45 |
+
"model_name": [
|
| 46 |
+
null,
|
| 47 |
+
null
|
| 48 |
+
],
|
| 49 |
+
"model_author": [
|
| 50 |
+
null,
|
| 51 |
+
null
|
| 52 |
+
],
|
| 53 |
+
"custom_dataset_info": [],
|
| 54 |
+
"quant_method": null,
|
| 55 |
+
"quant_bits": null,
|
| 56 |
+
"hqq_axis": null,
|
| 57 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 58 |
+
"bnb_4bit_quant_type": "nf4",
|
| 59 |
+
"bnb_4bit_use_double_quant": true,
|
| 60 |
+
"bnb_4bit_quant_storage": null,
|
| 61 |
+
"max_new_tokens": 64,
|
| 62 |
+
"temperature": 1.0,
|
| 63 |
+
"top_k": 50,
|
| 64 |
+
"top_p": 0.9,
|
| 65 |
+
"repetition_penalty": 1.1,
|
| 66 |
+
"num_beams": 1,
|
| 67 |
+
"stream": false,
|
| 68 |
+
"stop_words": [],
|
| 69 |
+
"logprobs": false,
|
| 70 |
+
"top_logprobs": null,
|
| 71 |
+
"ckpt_dir": null,
|
| 72 |
+
"load_dataset_config": null,
|
| 73 |
+
"lora_modules": [],
|
| 74 |
+
"tuner_backend": "peft",
|
| 75 |
+
"train_type": "lora",
|
| 76 |
+
"adapters": [],
|
| 77 |
+
"external_plugins": [
|
| 78 |
+
"/data/kdd/crag/cjz/ms-swift/examples/train/grpo/plugin/plugin.py"
|
| 79 |
+
],
|
| 80 |
+
"seed": 123,
|
| 81 |
+
"model_kwargs": {},
|
| 82 |
+
"load_args": false,
|
| 83 |
+
"load_data_args": false,
|
| 84 |
+
"use_hf": false,
|
| 85 |
+
"hub_token": null,
|
| 86 |
+
"custom_register_path": [],
|
| 87 |
+
"ignore_args_error": false,
|
| 88 |
+
"use_swift_lora": false,
|
| 89 |
+
"output_dir": "/data2/kdd/crag/cjz/agents/output/response/v63-20250514-105305",
|
| 90 |
+
"overwrite_output_dir": false,
|
| 91 |
+
"do_train": false,
|
| 92 |
+
"do_eval": false,
|
| 93 |
+
"do_predict": false,
|
| 94 |
+
"eval_strategy": "steps",
|
| 95 |
+
"prediction_loss_only": false,
|
| 96 |
+
"per_device_train_batch_size": 8,
|
| 97 |
+
"per_device_eval_batch_size": 8,
|
| 98 |
+
"per_gpu_train_batch_size": null,
|
| 99 |
+
"per_gpu_eval_batch_size": null,
|
| 100 |
+
"gradient_accumulation_steps": 16,
|
| 101 |
+
"eval_accumulation_steps": null,
|
| 102 |
+
"eval_delay": 0,
|
| 103 |
+
"torch_empty_cache_steps": null,
|
| 104 |
+
"learning_rate": 0.0002,
|
| 105 |
+
"weight_decay": 0.1,
|
| 106 |
+
"adam_beta1": 0.9,
|
| 107 |
+
"adam_beta2": 0.95,
|
| 108 |
+
"adam_epsilon": 1e-08,
|
| 109 |
+
"max_grad_norm": 0.5,
|
| 110 |
+
"num_train_epochs": 2.0,
|
| 111 |
+
"max_steps": -1,
|
| 112 |
+
"lr_scheduler_type": "cosine",
|
| 113 |
+
"lr_scheduler_kwargs": null,
|
| 114 |
+
"warmup_ratio": 0.05,
|
| 115 |
+
"warmup_steps": 0,
|
| 116 |
+
"log_level": "passive",
|
| 117 |
+
"log_level_replica": "warning",
|
| 118 |
+
"log_on_each_node": true,
|
| 119 |
+
"logging_dir": "/data2/kdd/crag/cjz/agents/output/response/v63-20250514-105305/runs",
|
| 120 |
+
"logging_strategy": "steps",
|
| 121 |
+
"logging_first_step": true,
|
| 122 |
+
"logging_steps": 1,
|
| 123 |
+
"logging_nan_inf_filter": true,
|
| 124 |
+
"save_strategy": "steps",
|
| 125 |
+
"save_steps": 5.0,
|
| 126 |
+
"save_total_limit": null,
|
| 127 |
+
"save_safetensors": true,
|
| 128 |
+
"save_on_each_node": false,
|
| 129 |
+
"save_only_model": false,
|
| 130 |
+
"restore_callback_states_from_checkpoint": false,
|
| 131 |
+
"no_cuda": false,
|
| 132 |
+
"use_cpu": false,
|
| 133 |
+
"use_mps_device": false,
|
| 134 |
+
"jit_mode_eval": false,
|
| 135 |
+
"use_ipex": false,
|
| 136 |
+
"bf16": true,
|
| 137 |
+
"fp16": false,
|
| 138 |
+
"fp16_opt_level": "O1",
|
| 139 |
+
"half_precision_backend": "auto",
|
| 140 |
+
"bf16_full_eval": false,
|
| 141 |
+
"fp16_full_eval": false,
|
| 142 |
+
"tf32": null,
|
| 143 |
+
"local_rank": 0,
|
| 144 |
+
"ddp_backend": null,
|
| 145 |
+
"tpu_num_cores": null,
|
| 146 |
+
"tpu_metrics_debug": false,
|
| 147 |
+
"debug": null,
|
| 148 |
+
"dataloader_drop_last": true,
|
| 149 |
+
"eval_steps": 500.0,
|
| 150 |
+
"dataloader_num_workers": 4,
|
| 151 |
+
"dataloader_prefetch_factor": null,
|
| 152 |
+
"past_index": -1,
|
| 153 |
+
"run_name": null,
|
| 154 |
+
"disable_tqdm": null,
|
| 155 |
+
"label_names": null,
|
| 156 |
+
"load_best_model_at_end": false,
|
| 157 |
+
"metric_for_best_model": "reward",
|
| 158 |
+
"greater_is_better": true,
|
| 159 |
+
"ignore_data_skip": false,
|
| 160 |
+
"fsdp": "",
|
| 161 |
+
"fsdp_min_num_params": 0,
|
| 162 |
+
"fsdp_config": null,
|
| 163 |
+
"tp_size": 0,
|
| 164 |
+
"fsdp_transformer_layer_cls_to_wrap": null,
|
| 165 |
+
"accelerator_config": {
|
| 166 |
+
"dispatch_batches": false
|
| 167 |
+
},
|
| 168 |
+
"deepspeed": {
|
| 169 |
+
"fp16": {
|
| 170 |
+
"enabled": "auto",
|
| 171 |
+
"loss_scale": 0,
|
| 172 |
+
"loss_scale_window": 1000,
|
| 173 |
+
"initial_scale_power": 16,
|
| 174 |
+
"hysteresis": 2,
|
| 175 |
+
"min_loss_scale": 1
|
| 176 |
+
},
|
| 177 |
+
"bf16": {
|
| 178 |
+
"enabled": "auto"
|
| 179 |
+
},
|
| 180 |
+
"zero_optimization": {
|
| 181 |
+
"stage": 3,
|
| 182 |
+
"offload_optimizer": {
|
| 183 |
+
"device": "none",
|
| 184 |
+
"pin_memory": true
|
| 185 |
+
},
|
| 186 |
+
"offload_param": {
|
| 187 |
+
"device": "none",
|
| 188 |
+
"pin_memory": true
|
| 189 |
+
},
|
| 190 |
+
"overlap_comm": false,
|
| 191 |
+
"contiguous_gradients": true,
|
| 192 |
+
"sub_group_size": 1000000000.0,
|
| 193 |
+
"reduce_bucket_size": "auto",
|
| 194 |
+
"zero_quantized_weights": false,
|
| 195 |
+
"zero_quantized_gradients": false,
|
| 196 |
+
"stage3_prefetch_bucket_size": "auto",
|
| 197 |
+
"stage3_param_persistence_threshold": "auto",
|
| 198 |
+
"stage3_max_live_parameters": 1000000000.0,
|
| 199 |
+
"stage3_max_reuse_distance": 1000000000.0,
|
| 200 |
+
"stage3_gather_16bit_weights_on_model_save": true
|
| 201 |
+
},
|
| 202 |
+
"gradient_accumulation_steps": "auto",
|
| 203 |
+
"gradient_clipping": "auto",
|
| 204 |
+
"steps_per_print": 2000,
|
| 205 |
+
"train_batch_size": "auto",
|
| 206 |
+
"train_micro_batch_size_per_gpu": "auto",
|
| 207 |
+
"wall_clock_breakdown": false
|
| 208 |
+
},
|
| 209 |
+
"label_smoothing_factor": 0.0,
|
| 210 |
+
"optim": "adamw_torch",
|
| 211 |
+
"optim_args": null,
|
| 212 |
+
"adafactor": false,
|
| 213 |
+
"group_by_length": false,
|
| 214 |
+
"length_column_name": "length",
|
| 215 |
+
"report_to": [
|
| 216 |
+
"tensorboard"
|
| 217 |
+
],
|
| 218 |
+
"ddp_find_unused_parameters": null,
|
| 219 |
+
"ddp_bucket_cap_mb": null,
|
| 220 |
+
"ddp_broadcast_buffers": null,
|
| 221 |
+
"dataloader_pin_memory": true,
|
| 222 |
+
"dataloader_persistent_workers": false,
|
| 223 |
+
"skip_memory_metrics": true,
|
| 224 |
+
"use_legacy_prediction_loop": false,
|
| 225 |
+
"push_to_hub": false,
|
| 226 |
+
"resume_from_checkpoint": "/data2/kdd/crag/cjz/agents/output/response/v55-20250513-012335/checkpoint-100",
|
| 227 |
+
"hub_model_id": null,
|
| 228 |
+
"hub_strategy": "every_save",
|
| 229 |
+
"hub_private_repo": null,
|
| 230 |
+
"hub_always_push": false,
|
| 231 |
+
"gradient_checkpointing": true,
|
| 232 |
+
"gradient_checkpointing_kwargs": null,
|
| 233 |
+
"include_inputs_for_metrics": false,
|
| 234 |
+
"include_for_metrics": [],
|
| 235 |
+
"eval_do_concat_batches": true,
|
| 236 |
+
"fp16_backend": "auto",
|
| 237 |
+
"push_to_hub_model_id": null,
|
| 238 |
+
"push_to_hub_organization": null,
|
| 239 |
+
"push_to_hub_token": null,
|
| 240 |
+
"mp_parameters": "",
|
| 241 |
+
"auto_find_batch_size": false,
|
| 242 |
+
"full_determinism": false,
|
| 243 |
+
"torchdynamo": null,
|
| 244 |
+
"ray_scope": "last",
|
| 245 |
+
"ddp_timeout": 1800,
|
| 246 |
+
"torch_compile": false,
|
| 247 |
+
"torch_compile_backend": null,
|
| 248 |
+
"torch_compile_mode": null,
|
| 249 |
+
"include_tokens_per_second": false,
|
| 250 |
+
"include_num_input_tokens_seen": false,
|
| 251 |
+
"neftune_noise_alpha": null,
|
| 252 |
+
"optim_target_modules": null,
|
| 253 |
+
"batch_eval_metrics": false,
|
| 254 |
+
"eval_on_start": false,
|
| 255 |
+
"use_liger_kernel": false,
|
| 256 |
+
"eval_use_gather_object": false,
|
| 257 |
+
"average_tokens_across_devices": false,
|
| 258 |
+
"sortish_sampler": false,
|
| 259 |
+
"predict_with_generate": false,
|
| 260 |
+
"generation_max_length": null,
|
| 261 |
+
"generation_num_beams": null,
|
| 262 |
+
"generation_config": null,
|
| 263 |
+
"check_model": true,
|
| 264 |
+
"acc_strategy": "token",
|
| 265 |
+
"train_dataloader_shuffle": true,
|
| 266 |
+
"metric_warmup_step": 0,
|
| 267 |
+
"fsdp_num": 1,
|
| 268 |
+
"acc_steps": 1,
|
| 269 |
+
"eval_use_evalscope": false,
|
| 270 |
+
"eval_datasets": [],
|
| 271 |
+
"eval_limit": null,
|
| 272 |
+
"eval_datasets_args": null,
|
| 273 |
+
"eval_generation_config": null,
|
| 274 |
+
"freeze_parameters": [
|
| 275 |
+
"multi_modal_projector"
|
| 276 |
+
],
|
| 277 |
+
"freeze_parameters_ratio": 0.0,
|
| 278 |
+
"trainable_parameters": [],
|
| 279 |
+
"freeze_llm": false,
|
| 280 |
+
"freeze_vit": false,
|
| 281 |
+
"freeze_aligner": true,
|
| 282 |
+
"target_modules": [
|
| 283 |
+
"all-linear"
|
| 284 |
+
],
|
| 285 |
+
"target_regex": null,
|
| 286 |
+
"modules_to_save": [],
|
| 287 |
+
"lora_rank": 8,
|
| 288 |
+
"lora_alpha": 32,
|
| 289 |
+
"lora_dropout": 0.05,
|
| 290 |
+
"lora_bias": "none",
|
| 291 |
+
"lora_dtype": null,
|
| 292 |
+
"lorap_lr_ratio": null,
|
| 293 |
+
"use_rslora": false,
|
| 294 |
+
"use_dora": false,
|
| 295 |
+
"lora_ga_batch_size": 2,
|
| 296 |
+
"lora_ga_iters": 2,
|
| 297 |
+
"lora_ga_max_length": 1024,
|
| 298 |
+
"lora_ga_direction": "ArB2r",
|
| 299 |
+
"lora_ga_scale": "stable",
|
| 300 |
+
"lora_ga_stable_gamma": 16,
|
| 301 |
+
"init_weights": true,
|
| 302 |
+
"fourier_n_frequency": 2000,
|
| 303 |
+
"fourier_scaling": 300.0,
|
| 304 |
+
"boft_block_size": 4,
|
| 305 |
+
"boft_block_num": 0,
|
| 306 |
+
"boft_n_butterfly_factor": 1,
|
| 307 |
+
"boft_dropout": 0.0,
|
| 308 |
+
"vera_rank": 256,
|
| 309 |
+
"vera_projection_prng_key": 0,
|
| 310 |
+
"vera_dropout": 0.0,
|
| 311 |
+
"vera_d_initial": 0.1,
|
| 312 |
+
"adapter_act": "gelu",
|
| 313 |
+
"adapter_length": 128,
|
| 314 |
+
"use_galore": false,
|
| 315 |
+
"galore_target_modules": null,
|
| 316 |
+
"galore_rank": 128,
|
| 317 |
+
"galore_update_proj_gap": 50,
|
| 318 |
+
"galore_scale": 1.0,
|
| 319 |
+
"galore_proj_type": "std",
|
| 320 |
+
"galore_optim_per_parameter": false,
|
| 321 |
+
"galore_with_embedding": false,
|
| 322 |
+
"galore_quantization": false,
|
| 323 |
+
"galore_proj_quant": false,
|
| 324 |
+
"galore_proj_bits": 4,
|
| 325 |
+
"galore_proj_group_size": 256,
|
| 326 |
+
"galore_cos_threshold": 0.4,
|
| 327 |
+
"galore_gamma_proj": 2,
|
| 328 |
+
"galore_queue_size": 5,
|
| 329 |
+
"adalora_target_r": 8,
|
| 330 |
+
"adalora_init_r": 12,
|
| 331 |
+
"adalora_tinit": 0,
|
| 332 |
+
"adalora_tfinal": 0,
|
| 333 |
+
"adalora_deltaT": 1,
|
| 334 |
+
"adalora_beta1": 0.85,
|
| 335 |
+
"adalora_beta2": 0.85,
|
| 336 |
+
"adalora_orth_reg_weight": 0.5,
|
| 337 |
+
"llamapro_num_new_blocks": 4,
|
| 338 |
+
"llamapro_num_groups": null,
|
| 339 |
+
"lisa_activated_layers": 0,
|
| 340 |
+
"lisa_step_interval": 20,
|
| 341 |
+
"reft_layer_key": null,
|
| 342 |
+
"reft_layers": null,
|
| 343 |
+
"reft_rank": 4,
|
| 344 |
+
"reft_intervention_type": "LoreftIntervention",
|
| 345 |
+
"reft_args": null,
|
| 346 |
+
"swanlab_token": null,
|
| 347 |
+
"swanlab_project": null,
|
| 348 |
+
"swanlab_workspace": null,
|
| 349 |
+
"swanlab_exp_name": null,
|
| 350 |
+
"swanlab_mode": "cloud",
|
| 351 |
+
"add_version": true,
|
| 352 |
+
"resume_only_model": false,
|
| 353 |
+
"create_checkpoint_symlink": false,
|
| 354 |
+
"packing": false,
|
| 355 |
+
"lazy_tokenize": true,
|
| 356 |
+
"loss_type": "grpo",
|
| 357 |
+
"optimizer": null,
|
| 358 |
+
"metric": null,
|
| 359 |
+
"zero_hpz_partition_size": null,
|
| 360 |
+
"reward_model": null,
|
| 361 |
+
"reward_adapters": [],
|
| 362 |
+
"reward_model_type": null,
|
| 363 |
+
"reward_model_revision": null,
|
| 364 |
+
"num_ppo_epochs": 4,
|
| 365 |
+
"whiten_rewards": false,
|
| 366 |
+
"kl_coef": 0.05,
|
| 367 |
+
"cliprange": 0.2,
|
| 368 |
+
"vf_coef": 0.1,
|
| 369 |
+
"cliprange_value": 0.2,
|
| 370 |
+
"gamma": 1.0,
|
| 371 |
+
"lam": 0.95,
|
| 372 |
+
"num_mini_batches": 1,
|
| 373 |
+
"local_rollout_forward_batch_size": 64,
|
| 374 |
+
"num_sample_generations": 10,
|
| 375 |
+
"response_length": 512,
|
| 376 |
+
"missing_eos_penalty": null,
|
| 377 |
+
"epsilon": 0.2,
|
| 378 |
+
"epsilon_high": 0.28,
|
| 379 |
+
"num_infer_workers": 1,
|
| 380 |
+
"vllm_device": [
|
| 381 |
+
"auto"
|
| 382 |
+
],
|
| 383 |
+
"vllm_gpu_memory_utilization": 0.8,
|
| 384 |
+
"vllm_max_model_len": 8192,
|
| 385 |
+
"vllm_max_num_seqs": 8,
|
| 386 |
+
"vllm_enforce_eager": false,
|
| 387 |
+
"vllm_limit_mm_per_prompt": null,
|
| 388 |
+
"vllm_enable_prefix_caching": true,
|
| 389 |
+
"cosine_min_len_value_wrong": -0.5,
|
| 390 |
+
"cosine_max_len_value_wrong": 0.0,
|
| 391 |
+
"cosine_min_len_value_correct": 1.0,
|
| 392 |
+
"cosine_max_len_value_correct": 0.5,
|
| 393 |
+
"cosine_max_len": null,
|
| 394 |
+
"repetition_n_grams": 3,
|
| 395 |
+
"repetition_max_penalty": -1.0,
|
| 396 |
+
"use_lmdeploy": false,
|
| 397 |
+
"lmdeploy_device": "auto",
|
| 398 |
+
"lmdeploy_session_len": null,
|
| 399 |
+
"lmdeploy_cache_max_entry_count": 0.8,
|
| 400 |
+
"async_generate": false,
|
| 401 |
+
"tensor_parallel_size": 1,
|
| 402 |
+
"sleep_level": 0,
|
| 403 |
+
"move_model_batches": null,
|
| 404 |
+
"offload_optimizer": false,
|
| 405 |
+
"offload_model": false,
|
| 406 |
+
"gc_collect_after_offload": false,
|
| 407 |
+
"multi_turn_func": null,
|
| 408 |
+
"dynamic_sample": false,
|
| 409 |
+
"max_resample_times": 3,
|
| 410 |
+
"overlong_filter": false,
|
| 411 |
+
"soft_max_length": null,
|
| 412 |
+
"soft_cache_length": 20,
|
| 413 |
+
"scale_rewards": true,
|
| 414 |
+
"wandb_log_unique_prompts": null,
|
| 415 |
+
"vllm_server_host": null,
|
| 416 |
+
"vllm_server_port": 8000,
|
| 417 |
+
"vllm_server_timeout": 240.0,
|
| 418 |
+
"num_generations": 8,
|
| 419 |
+
"max_completion_length": 50,
|
| 420 |
+
"ds3_gather_for_generation": true,
|
| 421 |
+
"reward_funcs": [
|
| 422 |
+
"Response_no_think_reward_1"
|
| 423 |
+
],
|
| 424 |
+
"reward_weights": null,
|
| 425 |
+
"log_completions": true,
|
| 426 |
+
"use_vllm": true,
|
| 427 |
+
"num_iterations": 2,
|
| 428 |
+
"rlhf_type": "grpo",
|
| 429 |
+
"ref_model": null,
|
| 430 |
+
"ref_model_type": null,
|
| 431 |
+
"ref_model_revision": null,
|
| 432 |
+
"beta": 0.01,
|
| 433 |
+
"label_smoothing": 0,
|
| 434 |
+
"rpo_alpha": 1.0,
|
| 435 |
+
"cpo_alpha": 1.0,
|
| 436 |
+
"simpo_gamma": 1,
|
| 437 |
+
"desirable_weight": 1.0,
|
| 438 |
+
"undesirable_weight": 1.0,
|
| 439 |
+
"center_rewards_coefficient": null,
|
| 440 |
+
"rank": 0,
|
| 441 |
+
"global_world_size": 1,
|
| 442 |
+
"local_world_size": 1,
|
| 443 |
+
"model_suffix": "Llama-3.2-11B-Vision-Instruct",
|
| 444 |
+
"model_info": "ModelInfo(model_type='llama3_2_vision', model_dir='/data1/chenjiazun/LLM/Llama-3.2-11B-Vision-Instruct', torch_dtype=torch.bfloat16, max_model_len=131072, quant_method=None, quant_bits=None, rope_scaling={'factor': 8.0, 'high_freq_factor': 4.0, 'low_freq_factor': 1.0, 'original_max_position_embeddings': 8192, 'rope_type': 'llama3'}, config=None, task_type='causal_lm', num_labels=None)",
|
| 445 |
+
"model_meta": "ModelMeta(model_type='llama3_2_vision', model_groups=[ModelGroup(models=[Model(ms_model_id='LLM-Research/Llama-3.2-11B-Vision-Instruct', hf_model_id='meta-llama/Llama-3.2-11B-Vision-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='LLM-Research/Llama-3.2-90B-Vision-Instruct', hf_model_id='meta-llama/Llama-3.2-90B-Vision-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='LLM-Research/Llama-3.2-11B-Vision', hf_model_id='meta-llama/Llama-3.2-11B-Vision', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='LLM-Research/Llama-3.2-90B-Vision', hf_model_id='meta-llama/Llama-3.2-90B-Vision', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='llama3_2_vision', get_function=<function get_model_tokenizer_llama3_2_vision at 0x7e26ea8c3d90>, model_arch='llama3_2_vision', architectures=['MllamaForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.45'], tags=[])",
|
| 446 |
+
"model_dir": "/data1/chenjiazun/LLM/Llama-3.2-11B-Vision-Instruct",
|
| 447 |
+
"hub": "<class 'swift.hub.hub.MSHub'>",
|
| 448 |
+
"evaluation_strategy": "steps",
|
| 449 |
+
"training_args": "GRPOConfig(output_dir='/data2/kdd/crag/cjz/agents/output/response/v63-20250514-105305', overwrite_output_dir=False, do_train=False, do_eval=True, do_predict=False, eval_strategy=<IntervalStrategy.STEPS: 'steps'>, prediction_loss_only=False, per_device_train_batch_size=8, per_device_eval_batch_size=8, per_gpu_train_batch_size=None, per_gpu_eval_batch_size=None, gradient_accumulation_steps=16, eval_accumulation_steps=None, eval_delay=0, torch_empty_cache_steps=None, learning_rate=0.0002, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, max_grad_norm=0.5, num_train_epochs=2.0, max_steps=-1, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_ratio=0.05, warmup_steps=0, log_level='passive', log_level_replica='warning', log_on_each_node=True, logging_dir='/data2/kdd/crag/cjz/agents/output/response/v63-20250514-105305/runs', logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_first_step=True, logging_steps=1, logging_nan_inf_filter=True, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=5, save_total_limit=None, save_safetensors=True, save_on_each_node=False, save_only_model=False, restore_callback_states_from_checkpoint=False, no_cuda=False, use_cpu=False, use_mps_device=False, seed=123, data_seed=123, jit_mode_eval=False, use_ipex=False, bf16=True, fp16=False, fp16_opt_level='O1', half_precision_backend='auto', bf16_full_eval=False, fp16_full_eval=False, tf32=None, local_rank=0, ddp_backend=None, tpu_num_cores=None, tpu_metrics_debug=False, debug=[], dataloader_drop_last=True, eval_steps=500, dataloader_num_workers=4, dataloader_prefetch_factor=10, past_index=-1, run_name='/data2/kdd/crag/cjz/agents/output/response/v63-20250514-105305', disable_tqdm=False, remove_unused_columns=False, label_names=None, load_best_model_at_end=False, metric_for_best_model='reward', greater_is_better=True, ignore_data_skip=False, fsdp=[], fsdp_min_num_params=0, fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, tp_size=0, fsdp_transformer_layer_cls_to_wrap=None, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 3, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'offload_param': {'device': 'none', 'pin_memory': True}, 'overlap_comm': False, 'contiguous_gradients': True, 'sub_group_size': 1000000000.0, 'reduce_bucket_size': 'auto', 'zero_quantized_weights': False, 'zero_quantized_gradients': False, 'stage3_prefetch_bucket_size': 0, 'stage3_param_persistence_threshold': 'auto', 'stage3_max_live_parameters': 1000000000.0, 'stage3_max_reuse_distance': 1000000000.0, 'stage3_gather_16bit_weights_on_model_save': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, label_smoothing_factor=0.0, optim=<OptimizerNames.ADAMW_TORCH: 'adamw_torch'>, optim_args=None, adafactor=False, group_by_length=False, length_column_name='length', report_to=['tensorboard'], ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, dataloader_pin_memory=True, dataloader_persistent_workers=False, skip_memory_metrics=True, use_legacy_prediction_loop=False, push_to_hub=False, resume_from_checkpoint='/data2/kdd/crag/cjz/agents/output/response/v55-20250513-012335/checkpoint-100', hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_token=None, hub_private_repo=None, hub_always_push=False, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, include_inputs_for_metrics=False, include_for_metrics=[], eval_do_concat_batches=True, fp16_backend='auto', push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=None, mp_parameters='', auto_find_batch_size=False, full_determinism=False, torchdynamo=None, ray_scope='last', ddp_timeout=1800, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, include_tokens_per_second=None, include_num_input_tokens_seen=None, neftune_noise_alpha=None, optim_target_modules=None, batch_eval_metrics=False, eval_on_start=False, use_liger_kernel=False, eval_use_gather_object=False, average_tokens_across_devices=None, model_init_kwargs=None, disable_dropout=False, max_prompt_length=512, num_generations=8, max_completion_length=50, ds3_gather_for_generation=True, shuffle_dataset=True, temperature=1.0, top_p=0.9, top_k=50, min_p=None, repetition_penalty=1.1, cache_implementation=None, use_vllm=True, vllm_server_host=None, vllm_server_port=8000, vllm_server_timeout=240.0, vllm_guided_decoding_regex=None, beta=0.01, num_iterations=2, epsilon=0.2, epsilon_high=0.28, reward_weights=None, scale_rewards=True, loss_type='grpo', mask_truncated_completions=False, sync_ref_model=False, ref_model_mixup_alpha=0.6, ref_model_sync_steps=512, use_liger_loss=False, log_completions=True, num_completions_to_print=None, wandb_log_unique_prompts=None, vllm_device=['auto'], vllm_gpu_memory_utilization=0.8, vllm_dtype=None, vllm_max_model_len=8192, vllm_enable_prefix_caching=True, check_model=True, acc_strategy='token', train_dataloader_shuffle=True, metric_warmup_step=0, fsdp_num=1, acc_steps=1, eval_use_evalscope=False, eval_datasets=[], eval_limit=None, eval_datasets_args=None, eval_generation_config=None, train_type='lora', optimizer=None, local_repo_path=None, galore_config=None, num_infer_workers=1, vllm_max_num_seqs=8, vllm_enforce_eager=False, vllm_limit_mm_per_prompt={}, cosine_min_len_value_wrong=-0.5, cosine_max_len_value_wrong=0.0, cosine_min_len_value_correct=1.0, cosine_max_len_value_correct=0.5, cosine_max_len=50, repetition_n_grams=3, repetition_max_penalty=-1.0, use_lmdeploy=False, lmdeploy_device='auto', lmdeploy_session_len=None, lmdeploy_cache_max_entry_count=0.8, async_generate=False, tensor_parallel_size=1, sleep_level=0, move_model_batches=None, offload_optimizer=False, offload_model=False, gc_collect_after_offload=False, multi_turn_func=None, dynamic_sample=False, max_resample_times=3, overlong_filter=False, soft_max_length=None, soft_cache_length=20, dataset_shuffle=True, stop_words=[])"
|
| 450 |
+
}
|
latest
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
global_step130
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:60cf7af1e877936bac8d9a6592c9375c51431a40757a09196bb041b48c14cfa4
|
| 3 |
+
size 14498
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a241e751a2cc669b60010ed509b4ce280ab2de58637775fdda76d0787b73239
|
| 3 |
+
size 1064
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2504 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 1.443213296398892,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 130,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"clip_ratio/high_max": 0.0,
|
| 14 |
+
"clip_ratio/high_mean": 0.0,
|
| 15 |
+
"clip_ratio/low_mean": 0.0,
|
| 16 |
+
"clip_ratio/low_min": 0.0,
|
| 17 |
+
"clip_ratio/region_mean": 0.0,
|
| 18 |
+
"completions/clipped_ratio": 0.0234375,
|
| 19 |
+
"completions/max_length": 101.0,
|
| 20 |
+
"completions/mean_length": 14.6328125,
|
| 21 |
+
"completions/min_length": 2.0,
|
| 22 |
+
"epoch": 0.0110803324099723,
|
| 23 |
+
"grad_norm": 1.5475432762019699,
|
| 24 |
+
"kl": 0.0,
|
| 25 |
+
"learning_rate": 2.2222222222222223e-05,
|
| 26 |
+
"loss": 5.587935447692871e-09,
|
| 27 |
+
"memory(GiB)": 77.28,
|
| 28 |
+
"reward": 0.25,
|
| 29 |
+
"reward_std": 0.7772719860076904,
|
| 30 |
+
"rewards/Response_no_think_reward/mean": 0.25,
|
| 31 |
+
"rewards/Response_no_think_reward/std": 1.4795188903808594,
|
| 32 |
+
"step": 1,
|
| 33 |
+
"train_speed(iter/s)": 0.001043
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"clip_ratio/high_max": 0.0,
|
| 37 |
+
"clip_ratio/high_mean": 0.0,
|
| 38 |
+
"clip_ratio/low_mean": 0.0,
|
| 39 |
+
"clip_ratio/low_min": 0.0,
|
| 40 |
+
"clip_ratio/region_mean": 0.0,
|
| 41 |
+
"epoch": 0.0221606648199446,
|
| 42 |
+
"grad_norm": 1.552548658770951,
|
| 43 |
+
"kl": 0.0,
|
| 44 |
+
"learning_rate": 4.4444444444444447e-05,
|
| 45 |
+
"loss": 5.587935447692871e-09,
|
| 46 |
+
"memory(GiB)": 78.06,
|
| 47 |
+
"step": 2,
|
| 48 |
+
"train_speed(iter/s)": 0.001737
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"clip_ratio/high_max": 0.0,
|
| 52 |
+
"clip_ratio/high_mean": 0.0,
|
| 53 |
+
"clip_ratio/low_mean": 0.0,
|
| 54 |
+
"clip_ratio/low_min": 0.0,
|
| 55 |
+
"clip_ratio/region_mean": 0.0,
|
| 56 |
+
"completions/clipped_ratio": 0.03125,
|
| 57 |
+
"completions/max_length": 101.0,
|
| 58 |
+
"completions/mean_length": 17.8828125,
|
| 59 |
+
"completions/min_length": 3.0,
|
| 60 |
+
"epoch": 0.0332409972299169,
|
| 61 |
+
"grad_norm": 1.291839688026328,
|
| 62 |
+
"kl": 0.0012106498143111821,
|
| 63 |
+
"learning_rate": 6.666666666666667e-05,
|
| 64 |
+
"loss": 0.0004578572406899184,
|
| 65 |
+
"memory(GiB)": 78.06,
|
| 66 |
+
"reward": 0.25,
|
| 67 |
+
"reward_std": 0.8250656127929688,
|
| 68 |
+
"rewards/Response_no_think_reward/mean": 0.25,
|
| 69 |
+
"rewards/Response_no_think_reward/std": 1.4634658098220825,
|
| 70 |
+
"step": 3,
|
| 71 |
+
"train_speed(iter/s)": 0.001451
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"clip_ratio/high_max": 0.004197790374746546,
|
| 75 |
+
"clip_ratio/high_mean": 0.004197790374746546,
|
| 76 |
+
"clip_ratio/low_mean": 0.012020835361909121,
|
| 77 |
+
"clip_ratio/low_min": 0.012020835361909121,
|
| 78 |
+
"clip_ratio/region_mean": 0.01621862585307099,
|
| 79 |
+
"epoch": 0.0443213296398892,
|
| 80 |
+
"grad_norm": 1.0301802967661304,
|
| 81 |
+
"kl": 0.005412253500253428,
|
| 82 |
+
"learning_rate": 8.888888888888889e-05,
|
| 83 |
+
"loss": -0.005840146914124489,
|
| 84 |
+
"memory(GiB)": 78.06,
|
| 85 |
+
"step": 4,
|
| 86 |
+
"train_speed(iter/s)": 0.001769
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"clip_ratio/high_max": 0.0,
|
| 90 |
+
"clip_ratio/high_mean": 0.0,
|
| 91 |
+
"clip_ratio/low_mean": 0.002237339736893773,
|
| 92 |
+
"clip_ratio/low_min": 0.002237339736893773,
|
| 93 |
+
"clip_ratio/region_mean": 0.002237339736893773,
|
| 94 |
+
"completions/clipped_ratio": 0.0078125,
|
| 95 |
+
"completions/max_length": 101.0,
|
| 96 |
+
"completions/mean_length": 12.5546875,
|
| 97 |
+
"completions/min_length": 3.0,
|
| 98 |
+
"epoch": 0.055401662049861494,
|
| 99 |
+
"grad_norm": 2.027395284218205,
|
| 100 |
+
"kl": 0.014184385421685874,
|
| 101 |
+
"learning_rate": 0.00011111111111111112,
|
| 102 |
+
"loss": 0.0008896891959011555,
|
| 103 |
+
"memory(GiB)": 78.06,
|
| 104 |
+
"reward": 0.8515625,
|
| 105 |
+
"reward_std": 0.5118520259857178,
|
| 106 |
+
"rewards/Response_no_think_reward/mean": 0.8515625,
|
| 107 |
+
"rewards/Response_no_think_reward/std": 1.4147136211395264,
|
| 108 |
+
"step": 5,
|
| 109 |
+
"train_speed(iter/s)": 0.00164
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"clip_ratio/high_max": 0.013844632252585143,
|
| 113 |
+
"clip_ratio/high_mean": 0.013844632252585143,
|
| 114 |
+
"clip_ratio/low_mean": 0.043920744908973575,
|
| 115 |
+
"clip_ratio/low_min": 0.043920744908973575,
|
| 116 |
+
"clip_ratio/region_mean": 0.05776537861675024,
|
| 117 |
+
"epoch": 0.0664819944598338,
|
| 118 |
+
"grad_norm": 6.212168397356187,
|
| 119 |
+
"kl": 1.7603500080003869,
|
| 120 |
+
"learning_rate": 0.00013333333333333334,
|
| 121 |
+
"loss": 0.01842591166496277,
|
| 122 |
+
"memory(GiB)": 79.98,
|
| 123 |
+
"step": 6,
|
| 124 |
+
"train_speed(iter/s)": 0.00185
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"clip_ratio/high_max": 0.0,
|
| 128 |
+
"clip_ratio/high_mean": 0.0,
|
| 129 |
+
"clip_ratio/low_mean": 0.001959657238330692,
|
| 130 |
+
"clip_ratio/low_min": 0.001959657238330692,
|
| 131 |
+
"clip_ratio/region_mean": 0.001959657238330692,
|
| 132 |
+
"completions/clipped_ratio": 0.0,
|
| 133 |
+
"completions/max_length": 70.0,
|
| 134 |
+
"completions/mean_length": 10.359375,
|
| 135 |
+
"completions/min_length": 2.0,
|
| 136 |
+
"epoch": 0.07756232686980609,
|
| 137 |
+
"grad_norm": 3.537674599924329,
|
| 138 |
+
"kl": 0.33545287084416486,
|
| 139 |
+
"learning_rate": 0.00015555555555555556,
|
| 140 |
+
"loss": 0.0018048422643914819,
|
| 141 |
+
"memory(GiB)": 79.98,
|
| 142 |
+
"reward": 0.53125,
|
| 143 |
+
"reward_std": 0.6121620535850525,
|
| 144 |
+
"rewards/Response_no_think_reward/mean": 0.53125,
|
| 145 |
+
"rewards/Response_no_think_reward/std": 1.3685873746871948,
|
| 146 |
+
"step": 7,
|
| 147 |
+
"train_speed(iter/s)": 0.001773
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"clip_ratio/high_max": 0.014042179333046079,
|
| 151 |
+
"clip_ratio/high_mean": 0.014042179333046079,
|
| 152 |
+
"clip_ratio/low_mean": 0.03925089433323592,
|
| 153 |
+
"clip_ratio/low_min": 0.03925089433323592,
|
| 154 |
+
"clip_ratio/region_mean": 0.053293074015527964,
|
| 155 |
+
"epoch": 0.0886426592797784,
|
| 156 |
+
"grad_norm": 2.64201486377639,
|
| 157 |
+
"kl": 1.295830228133127,
|
| 158 |
+
"learning_rate": 0.00017777777777777779,
|
| 159 |
+
"loss": 0.011679837480187416,
|
| 160 |
+
"memory(GiB)": 79.98,
|
| 161 |
+
"step": 8,
|
| 162 |
+
"train_speed(iter/s)": 0.001932
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"clip_ratio/high_max": 0.0011001251987181604,
|
| 166 |
+
"clip_ratio/high_mean": 0.0011001251987181604,
|
| 167 |
+
"clip_ratio/low_mean": 0.008499634364852682,
|
| 168 |
+
"clip_ratio/low_min": 0.008499634364852682,
|
| 169 |
+
"clip_ratio/region_mean": 0.009599759563570842,
|
| 170 |
+
"completions/clipped_ratio": 0.0,
|
| 171 |
+
"completions/max_length": 62.0,
|
| 172 |
+
"completions/mean_length": 11.4375,
|
| 173 |
+
"completions/min_length": 2.0,
|
| 174 |
+
"epoch": 0.0997229916897507,
|
| 175 |
+
"grad_norm": 4.1929387792795065,
|
| 176 |
+
"kl": 2.5014250652166083,
|
| 177 |
+
"learning_rate": 0.0002,
|
| 178 |
+
"loss": 0.016260012984275818,
|
| 179 |
+
"memory(GiB)": 79.98,
|
| 180 |
+
"reward": 0.2734375,
|
| 181 |
+
"reward_std": 0.8412302732467651,
|
| 182 |
+
"rewards/Response_no_think_reward/mean": 0.2734375,
|
| 183 |
+
"rewards/Response_no_think_reward/std": 1.384474277496338,
|
| 184 |
+
"step": 9,
|
| 185 |
+
"train_speed(iter/s)": 0.001887
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"clip_ratio/high_max": 0.02779325417941436,
|
| 189 |
+
"clip_ratio/high_mean": 0.02779325417941436,
|
| 190 |
+
"clip_ratio/low_mean": 0.07855538238072768,
|
| 191 |
+
"clip_ratio/low_min": 0.07855538238072768,
|
| 192 |
+
"clip_ratio/region_mean": 0.10634863609448075,
|
| 193 |
+
"epoch": 0.11080332409972299,
|
| 194 |
+
"grad_norm": 990.7609804346584,
|
| 195 |
+
"kl": 1.6530181597918272,
|
| 196 |
+
"learning_rate": 0.00019998312416333227,
|
| 197 |
+
"loss": 1.336751103401184,
|
| 198 |
+
"memory(GiB)": 79.98,
|
| 199 |
+
"step": 10,
|
| 200 |
+
"train_speed(iter/s)": 0.002015
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"clip_ratio/high_max": 0.001354054023977369,
|
| 204 |
+
"clip_ratio/high_mean": 0.001354054023977369,
|
| 205 |
+
"clip_ratio/low_mean": 0.004174399145995267,
|
| 206 |
+
"clip_ratio/low_min": 0.004174399145995267,
|
| 207 |
+
"clip_ratio/region_mean": 0.005528453169972636,
|
| 208 |
+
"completions/clipped_ratio": 0.0546875,
|
| 209 |
+
"completions/max_length": 101.0,
|
| 210 |
+
"completions/mean_length": 26.453125,
|
| 211 |
+
"completions/min_length": 3.0,
|
| 212 |
+
"epoch": 0.12188365650969529,
|
| 213 |
+
"grad_norm": 1.115304569915306,
|
| 214 |
+
"kl": 0.3547552889212966,
|
| 215 |
+
"learning_rate": 0.00019993250234920636,
|
| 216 |
+
"loss": 0.005897670052945614,
|
| 217 |
+
"memory(GiB)": 79.98,
|
| 218 |
+
"reward": 0.5234375,
|
| 219 |
+
"reward_std": 0.876558780670166,
|
| 220 |
+
"rewards/Response_no_think_reward/mean": 0.5234375,
|
| 221 |
+
"rewards/Response_no_think_reward/std": 1.4251114130020142,
|
| 222 |
+
"step": 11,
|
| 223 |
+
"train_speed(iter/s)": 0.001964
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"clip_ratio/high_max": 0.02347446102066897,
|
| 227 |
+
"clip_ratio/high_mean": 0.02347446102066897,
|
| 228 |
+
"clip_ratio/low_mean": 0.0428259114123648,
|
| 229 |
+
"clip_ratio/low_min": 0.0428259114123648,
|
| 230 |
+
"clip_ratio/region_mean": 0.06630037224385887,
|
| 231 |
+
"epoch": 0.1329639889196676,
|
| 232 |
+
"grad_norm": 2.183295109715178,
|
| 233 |
+
"kl": 0.6218942860141397,
|
| 234 |
+
"learning_rate": 0.00019984815164333163,
|
| 235 |
+
"loss": 0.007074224762618542,
|
| 236 |
+
"memory(GiB)": 79.98,
|
| 237 |
+
"step": 12,
|
| 238 |
+
"train_speed(iter/s)": 0.00207
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"clip_ratio/high_max": 0.0,
|
| 242 |
+
"clip_ratio/high_mean": 0.0,
|
| 243 |
+
"clip_ratio/low_mean": 0.001206563669256866,
|
| 244 |
+
"clip_ratio/low_min": 0.001206563669256866,
|
| 245 |
+
"clip_ratio/region_mean": 0.001206563669256866,
|
| 246 |
+
"completions/clipped_ratio": 0.0,
|
| 247 |
+
"completions/max_length": 83.0,
|
| 248 |
+
"completions/mean_length": 14.0234375,
|
| 249 |
+
"completions/min_length": 4.0,
|
| 250 |
+
"epoch": 0.1440443213296399,
|
| 251 |
+
"grad_norm": 1.0424369292135751,
|
| 252 |
+
"kl": 0.40942037590866676,
|
| 253 |
+
"learning_rate": 0.00019973010051548275,
|
| 254 |
+
"loss": 0.004908258095383644,
|
| 255 |
+
"memory(GiB)": 79.98,
|
| 256 |
+
"reward": 0.5625,
|
| 257 |
+
"reward_std": 0.7038782835006714,
|
| 258 |
+
"rewards/Response_no_think_reward/mean": 0.5625,
|
| 259 |
+
"rewards/Response_no_think_reward/std": 1.2720495462417603,
|
| 260 |
+
"step": 13,
|
| 261 |
+
"train_speed(iter/s)": 0.002042
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"clip_ratio/high_max": 0.017414433998055756,
|
| 265 |
+
"clip_ratio/high_mean": 0.017414433998055756,
|
| 266 |
+
"clip_ratio/low_mean": 0.028266766399610788,
|
| 267 |
+
"clip_ratio/low_min": 0.028266766399610788,
|
| 268 |
+
"clip_ratio/region_mean": 0.04568120092153549,
|
| 269 |
+
"epoch": 0.15512465373961218,
|
| 270 |
+
"grad_norm": 0.650161789807741,
|
| 271 |
+
"kl": 0.4890678570009186,
|
| 272 |
+
"learning_rate": 0.00019957838880989078,
|
| 273 |
+
"loss": -0.0075666941702365875,
|
| 274 |
+
"memory(GiB)": 79.98,
|
| 275 |
+
"step": 14,
|
| 276 |
+
"train_speed(iter/s)": 0.002135
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"clip_ratio/high_max": 0.0018629207770572975,
|
| 280 |
+
"clip_ratio/high_mean": 0.0018629207770572975,
|
| 281 |
+
"clip_ratio/low_mean": 0.0069178942940197885,
|
| 282 |
+
"clip_ratio/low_min": 0.0069178942940197885,
|
| 283 |
+
"clip_ratio/region_mean": 0.00878081505652517,
|
| 284 |
+
"completions/clipped_ratio": 0.0078125,
|
| 285 |
+
"completions/max_length": 101.0,
|
| 286 |
+
"completions/mean_length": 17.1953125,
|
| 287 |
+
"completions/min_length": 4.0,
|
| 288 |
+
"epoch": 0.16620498614958448,
|
| 289 |
+
"grad_norm": 2.2362713456466605,
|
| 290 |
+
"kl": 0.23374600778333843,
|
| 291 |
+
"learning_rate": 0.00019939306773179497,
|
| 292 |
+
"loss": 0.0009349192259833217,
|
| 293 |
+
"memory(GiB)": 79.98,
|
| 294 |
+
"reward": 0.6875,
|
| 295 |
+
"reward_std": 0.8529120683670044,
|
| 296 |
+
"rewards/Response_no_think_reward/mean": 0.6875,
|
| 297 |
+
"rewards/Response_no_think_reward/std": 1.3265905380249023,
|
| 298 |
+
"step": 15,
|
| 299 |
+
"train_speed(iter/s)": 0.002091
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"clip_ratio/high_max": 0.03511151310522109,
|
| 303 |
+
"clip_ratio/high_mean": 0.03511151310522109,
|
| 304 |
+
"clip_ratio/low_mean": 0.028677020454779267,
|
| 305 |
+
"clip_ratio/low_min": 0.028677020454779267,
|
| 306 |
+
"clip_ratio/region_mean": 0.06378853344358504,
|
| 307 |
+
"epoch": 0.1772853185595568,
|
| 308 |
+
"grad_norm": 3.3548685452904934,
|
| 309 |
+
"kl": 0.25148704896855634,
|
| 310 |
+
"learning_rate": 0.00019917419983016025,
|
| 311 |
+
"loss": 0.017475975677371025,
|
| 312 |
+
"memory(GiB)": 79.98,
|
| 313 |
+
"step": 16,
|
| 314 |
+
"train_speed(iter/s)": 0.002171
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"clip_ratio/high_max": 0.000469924823846668,
|
| 318 |
+
"clip_ratio/high_mean": 0.000469924823846668,
|
| 319 |
+
"clip_ratio/low_mean": 0.0014688223309349269,
|
| 320 |
+
"clip_ratio/low_min": 0.0014688223309349269,
|
| 321 |
+
"clip_ratio/region_mean": 0.0019387471547815949,
|
| 322 |
+
"completions/clipped_ratio": 0.0078125,
|
| 323 |
+
"completions/max_length": 101.0,
|
| 324 |
+
"completions/mean_length": 18.2265625,
|
| 325 |
+
"completions/min_length": 6.0,
|
| 326 |
+
"epoch": 0.1883656509695291,
|
| 327 |
+
"grad_norm": 0.6996032916694407,
|
| 328 |
+
"kl": 0.138063008276049,
|
| 329 |
+
"learning_rate": 0.00019892185897656578,
|
| 330 |
+
"loss": 0.0021217623725533485,
|
| 331 |
+
"memory(GiB)": 79.98,
|
| 332 |
+
"reward": 0.5,
|
| 333 |
+
"reward_std": 0.5098158717155457,
|
| 334 |
+
"rewards/Response_no_think_reward/mean": 0.5,
|
| 335 |
+
"rewards/Response_no_think_reward/std": 1.4086347818374634,
|
| 336 |
+
"step": 17,
|
| 337 |
+
"train_speed(iter/s)": 0.002127
|
| 338 |
+
},
|
| 339 |
+
{
|
| 340 |
+
"clip_ratio/high_max": 0.01048101403284818,
|
| 341 |
+
"clip_ratio/high_mean": 0.01048101403284818,
|
| 342 |
+
"clip_ratio/low_mean": 0.028478411200921983,
|
| 343 |
+
"clip_ratio/low_min": 0.028478411200921983,
|
| 344 |
+
"clip_ratio/region_mean": 0.03895942587405443,
|
| 345 |
+
"epoch": 0.1994459833795014,
|
| 346 |
+
"grad_norm": 0.7437416506383379,
|
| 347 |
+
"kl": 0.2438321103884391,
|
| 348 |
+
"learning_rate": 0.00019863613034027224,
|
| 349 |
+
"loss": -0.007616878021508455,
|
| 350 |
+
"memory(GiB)": 79.98,
|
| 351 |
+
"step": 18,
|
| 352 |
+
"train_speed(iter/s)": 0.002199
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"clip_ratio/high_max": 0.0,
|
| 356 |
+
"clip_ratio/high_mean": 0.0,
|
| 357 |
+
"clip_ratio/low_mean": 0.0009733444603625685,
|
| 358 |
+
"clip_ratio/low_min": 0.0009733444603625685,
|
| 359 |
+
"clip_ratio/region_mean": 0.0009733444603625685,
|
| 360 |
+
"completions/clipped_ratio": 0.0078125,
|
| 361 |
+
"completions/max_length": 101.0,
|
| 362 |
+
"completions/mean_length": 20.703125,
|
| 363 |
+
"completions/min_length": 6.0,
|
| 364 |
+
"epoch": 0.21052631578947367,
|
| 365 |
+
"grad_norm": 0.8303775863701255,
|
| 366 |
+
"kl": 0.09481111937202513,
|
| 367 |
+
"learning_rate": 0.0001983171103594755,
|
| 368 |
+
"loss": 0.0022672354243695736,
|
| 369 |
+
"memory(GiB)": 79.98,
|
| 370 |
+
"reward": -0.078125,
|
| 371 |
+
"reward_std": 0.6301807761192322,
|
| 372 |
+
"rewards/Response_no_think_reward/mean": -0.078125,
|
| 373 |
+
"rewards/Response_no_think_reward/std": 1.3666982650756836,
|
| 374 |
+
"step": 19,
|
| 375 |
+
"train_speed(iter/s)": 0.002152
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"clip_ratio/high_max": 0.007191620621597394,
|
| 379 |
+
"clip_ratio/high_mean": 0.007191620621597394,
|
| 380 |
+
"clip_ratio/low_mean": 0.03550086636096239,
|
| 381 |
+
"clip_ratio/low_min": 0.03550086636096239,
|
| 382 |
+
"clip_ratio/region_mean": 0.0426924874773249,
|
| 383 |
+
"epoch": 0.22160664819944598,
|
| 384 |
+
"grad_norm": 1.7213808531720889,
|
| 385 |
+
"kl": 1.6638920252444223,
|
| 386 |
+
"learning_rate": 0.0001979649067087574,
|
| 387 |
+
"loss": 0.006861768197268248,
|
| 388 |
+
"memory(GiB)": 79.98,
|
| 389 |
+
"step": 20,
|
| 390 |
+
"train_speed(iter/s)": 0.002217
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"clip_ratio/high_max": 0.0004032258002553135,
|
| 394 |
+
"clip_ratio/high_mean": 0.0004032258002553135,
|
| 395 |
+
"clip_ratio/low_mean": 0.003614576125983149,
|
| 396 |
+
"clip_ratio/low_min": 0.003614576125983149,
|
| 397 |
+
"clip_ratio/region_mean": 0.004017801926238462,
|
| 398 |
+
"completions/clipped_ratio": 0.0078125,
|
| 399 |
+
"completions/max_length": 101.0,
|
| 400 |
+
"completions/mean_length": 17.390625,
|
| 401 |
+
"completions/min_length": 6.0,
|
| 402 |
+
"epoch": 0.23268698060941828,
|
| 403 |
+
"grad_norm": 1.5808946015405074,
|
| 404 |
+
"kl": 0.1687323283404112,
|
| 405 |
+
"learning_rate": 0.00019757963826274357,
|
| 406 |
+
"loss": 0.003855248913168907,
|
| 407 |
+
"memory(GiB)": 79.98,
|
| 408 |
+
"reward": 0.5546875,
|
| 409 |
+
"reward_std": 0.8033354878425598,
|
| 410 |
+
"rewards/Response_no_think_reward/mean": 0.5546875,
|
| 411 |
+
"rewards/Response_no_think_reward/std": 1.3328590393066406,
|
| 412 |
+
"step": 21,
|
| 413 |
+
"train_speed(iter/s)": 0.002189
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"clip_ratio/high_max": 0.012200821249280125,
|
| 417 |
+
"clip_ratio/high_mean": 0.012200821249280125,
|
| 418 |
+
"clip_ratio/low_mean": 0.04007338697556406,
|
| 419 |
+
"clip_ratio/low_min": 0.04007338697556406,
|
| 420 |
+
"clip_ratio/region_mean": 0.05227420857409015,
|
| 421 |
+
"epoch": 0.24376731301939059,
|
| 422 |
+
"grad_norm": 0.8850881940917656,
|
| 423 |
+
"kl": 0.48105100472457707,
|
| 424 |
+
"learning_rate": 0.0001971614350559814,
|
| 425 |
+
"loss": -0.004205920733511448,
|
| 426 |
+
"memory(GiB)": 79.98,
|
| 427 |
+
"step": 22,
|
| 428 |
+
"train_speed(iter/s)": 0.002248
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"clip_ratio/high_max": 0.0008194574620574713,
|
| 432 |
+
"clip_ratio/high_mean": 0.0008194574620574713,
|
| 433 |
+
"clip_ratio/low_mean": 0.000811688310932368,
|
| 434 |
+
"clip_ratio/low_min": 0.000811688310932368,
|
| 435 |
+
"clip_ratio/region_mean": 0.0016311457729898393,
|
| 436 |
+
"completions/clipped_ratio": 0.0078125,
|
| 437 |
+
"completions/max_length": 101.0,
|
| 438 |
+
"completions/mean_length": 15.28125,
|
| 439 |
+
"completions/min_length": 6.0,
|
| 440 |
+
"epoch": 0.2548476454293629,
|
| 441 |
+
"grad_norm": 0.6954813297644332,
|
| 442 |
+
"kl": 0.11993603070732206,
|
| 443 |
+
"learning_rate": 0.0001967104382390511,
|
| 444 |
+
"loss": 0.001324990182183683,
|
| 445 |
+
"memory(GiB)": 79.98,
|
| 446 |
+
"reward": 0.78125,
|
| 447 |
+
"reward_std": 0.5077463984489441,
|
| 448 |
+
"rewards/Response_no_think_reward/mean": 0.78125,
|
| 449 |
+
"rewards/Response_no_think_reward/std": 1.3970582485198975,
|
| 450 |
+
"step": 23,
|
| 451 |
+
"train_speed(iter/s)": 0.002219
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"clip_ratio/high_max": 0.014543175988364965,
|
| 455 |
+
"clip_ratio/high_mean": 0.014543175988364965,
|
| 456 |
+
"clip_ratio/low_mean": 0.027941336738877,
|
| 457 |
+
"clip_ratio/low_min": 0.027941336738877,
|
| 458 |
+
"clip_ratio/region_mean": 0.04248451231978834,
|
| 459 |
+
"epoch": 0.2659279778393352,
|
| 460 |
+
"grad_norm": 0.46891914008721675,
|
| 461 |
+
"kl": 0.11759324668673798,
|
| 462 |
+
"learning_rate": 0.00019622680003092503,
|
| 463 |
+
"loss": -0.009693928062915802,
|
| 464 |
+
"memory(GiB)": 79.98,
|
| 465 |
+
"step": 24,
|
| 466 |
+
"train_speed(iter/s)": 0.002273
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"clip_ratio/high_max": 0.0006530559621751308,
|
| 470 |
+
"clip_ratio/high_mean": 0.0006530559621751308,
|
| 471 |
+
"clip_ratio/low_mean": 0.0019866162620019168,
|
| 472 |
+
"clip_ratio/low_min": 0.0019866162620019168,
|
| 473 |
+
"clip_ratio/region_mean": 0.002639672253280878,
|
| 474 |
+
"completions/clipped_ratio": 0.0078125,
|
| 475 |
+
"completions/max_length": 101.0,
|
| 476 |
+
"completions/mean_length": 12.9609375,
|
| 477 |
+
"completions/min_length": 4.0,
|
| 478 |
+
"epoch": 0.2770083102493075,
|
| 479 |
+
"grad_norm": 1.333182473379268,
|
| 480 |
+
"kl": 0.22027045290451497,
|
| 481 |
+
"learning_rate": 0.00019571068366759143,
|
| 482 |
+
"loss": 0.0026610023342072964,
|
| 483 |
+
"memory(GiB)": 79.98,
|
| 484 |
+
"reward": 0.4765625,
|
| 485 |
+
"reward_std": 0.8257243633270264,
|
| 486 |
+
"rewards/Response_no_think_reward/mean": 0.4765625,
|
| 487 |
+
"rewards/Response_no_think_reward/std": 1.3219220638275146,
|
| 488 |
+
"step": 25,
|
| 489 |
+
"train_speed(iter/s)": 0.002248
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"clip_ratio/high_max": 0.02311275008833036,
|
| 493 |
+
"clip_ratio/high_mean": 0.02311275008833036,
|
| 494 |
+
"clip_ratio/low_mean": 0.03803046100074425,
|
| 495 |
+
"clip_ratio/low_min": 0.03803046100074425,
|
| 496 |
+
"clip_ratio/region_mean": 0.06114321056520566,
|
| 497 |
+
"epoch": 0.2880886426592798,
|
| 498 |
+
"grad_norm": 0.7053648057917502,
|
| 499 |
+
"kl": 0.2949459235333052,
|
| 500 |
+
"learning_rate": 0.0001951622633469592,
|
| 501 |
+
"loss": -0.015641074627637863,
|
| 502 |
+
"memory(GiB)": 79.98,
|
| 503 |
+
"step": 26,
|
| 504 |
+
"train_speed(iter/s)": 0.002297
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"clip_ratio/high_max": 0.0004111842135898769,
|
| 508 |
+
"clip_ratio/high_mean": 0.0004111842135898769,
|
| 509 |
+
"clip_ratio/low_mean": 0.0027908546617254615,
|
| 510 |
+
"clip_ratio/low_min": 0.0027908546617254615,
|
| 511 |
+
"clip_ratio/region_mean": 0.0032020388753153384,
|
| 512 |
+
"completions/clipped_ratio": 0.0,
|
| 513 |
+
"completions/max_length": 85.0,
|
| 514 |
+
"completions/mean_length": 16.875,
|
| 515 |
+
"completions/min_length": 4.0,
|
| 516 |
+
"epoch": 0.29916897506925205,
|
| 517 |
+
"grad_norm": 1.544635006921068,
|
| 518 |
+
"kl": 0.20233443519100547,
|
| 519 |
+
"learning_rate": 0.00019458172417006347,
|
| 520 |
+
"loss": 0.0030936466064304113,
|
| 521 |
+
"memory(GiB)": 79.98,
|
| 522 |
+
"reward": 0.6875,
|
| 523 |
+
"reward_std": 0.9086803197860718,
|
| 524 |
+
"rewards/Response_no_think_reward/mean": 0.6875,
|
| 525 |
+
"rewards/Response_no_think_reward/std": 1.3265905380249023,
|
| 526 |
+
"step": 27,
|
| 527 |
+
"train_speed(iter/s)": 0.002272
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"clip_ratio/high_max": 0.03639351949095726,
|
| 531 |
+
"clip_ratio/high_mean": 0.03639351949095726,
|
| 532 |
+
"clip_ratio/low_mean": 0.024818695266731083,
|
| 533 |
+
"clip_ratio/low_min": 0.024818695266731083,
|
| 534 |
+
"clip_ratio/region_mean": 0.06121221440844238,
|
| 535 |
+
"epoch": 0.31024930747922436,
|
| 536 |
+
"grad_norm": 2.0012174050572105,
|
| 537 |
+
"kl": 0.24312824057415128,
|
| 538 |
+
"learning_rate": 0.00019396926207859084,
|
| 539 |
+
"loss": -0.008060472086071968,
|
| 540 |
+
"memory(GiB)": 79.98,
|
| 541 |
+
"step": 28,
|
| 542 |
+
"train_speed(iter/s)": 0.002319
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"clip_ratio/high_max": 0.0007382866751868278,
|
| 546 |
+
"clip_ratio/high_mean": 0.0007382866751868278,
|
| 547 |
+
"clip_ratio/low_mean": 0.0024169938114937395,
|
| 548 |
+
"clip_ratio/low_min": 0.0024169938114937395,
|
| 549 |
+
"clip_ratio/region_mean": 0.0031552804866805673,
|
| 550 |
+
"completions/clipped_ratio": 0.0234375,
|
| 551 |
+
"completions/max_length": 101.0,
|
| 552 |
+
"completions/mean_length": 23.4140625,
|
| 553 |
+
"completions/min_length": 3.0,
|
| 554 |
+
"epoch": 0.32132963988919666,
|
| 555 |
+
"grad_norm": 1.0333627248580772,
|
| 556 |
+
"kl": 0.19911292963661253,
|
| 557 |
+
"learning_rate": 0.0001933250837887457,
|
| 558 |
+
"loss": 0.004103388637304306,
|
| 559 |
+
"memory(GiB)": 79.98,
|
| 560 |
+
"reward": 0.6640625,
|
| 561 |
+
"reward_std": 0.7995060086250305,
|
| 562 |
+
"rewards/Response_no_think_reward/mean": 0.6640625,
|
| 563 |
+
"rewards/Response_no_think_reward/std": 1.4323447942733765,
|
| 564 |
+
"step": 29,
|
| 565 |
+
"train_speed(iter/s)": 0.002285
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"clip_ratio/high_max": 0.021052728639915586,
|
| 569 |
+
"clip_ratio/high_mean": 0.021052728639915586,
|
| 570 |
+
"clip_ratio/low_mean": 0.03886253567179665,
|
| 571 |
+
"clip_ratio/low_min": 0.03886253567179665,
|
| 572 |
+
"clip_ratio/region_mean": 0.05991526402067393,
|
| 573 |
+
"epoch": 0.33240997229916897,
|
| 574 |
+
"grad_norm": 0.4744156832935017,
|
| 575 |
+
"kl": 0.2653088476508856,
|
| 576 |
+
"learning_rate": 0.00019264940672148018,
|
| 577 |
+
"loss": -0.014724130742251873,
|
| 578 |
+
"memory(GiB)": 79.98,
|
| 579 |
+
"step": 30,
|
| 580 |
+
"train_speed(iter/s)": 0.002328
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"clip_ratio/high_max": 0.0008336337341461331,
|
| 584 |
+
"clip_ratio/high_mean": 0.0008336337341461331,
|
| 585 |
+
"clip_ratio/low_mean": 0.0028033541166223586,
|
| 586 |
+
"clip_ratio/low_min": 0.0028033541166223586,
|
| 587 |
+
"clip_ratio/region_mean": 0.0036369878507684916,
|
| 588 |
+
"completions/clipped_ratio": 0.0078125,
|
| 589 |
+
"completions/max_length": 101.0,
|
| 590 |
+
"completions/mean_length": 17.8125,
|
| 591 |
+
"completions/min_length": 3.0,
|
| 592 |
+
"epoch": 0.34349030470914127,
|
| 593 |
+
"grad_norm": 1.3971369328491496,
|
| 594 |
+
"kl": 0.3057649952825159,
|
| 595 |
+
"learning_rate": 0.0001919424589291108,
|
| 596 |
+
"loss": 0.0056169466115534306,
|
| 597 |
+
"memory(GiB)": 80.03,
|
| 598 |
+
"reward": 0.4296875,
|
| 599 |
+
"reward_std": 0.8740850687026978,
|
| 600 |
+
"rewards/Response_no_think_reward/mean": 0.4296875,
|
| 601 |
+
"rewards/Response_no_think_reward/std": 1.3555577993392944,
|
| 602 |
+
"step": 31,
|
| 603 |
+
"train_speed(iter/s)": 0.002304
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"clip_ratio/high_max": 0.013716876972466707,
|
| 607 |
+
"clip_ratio/high_mean": 0.013716876972466707,
|
| 608 |
+
"clip_ratio/low_mean": 0.05669466912513599,
|
| 609 |
+
"clip_ratio/low_min": 0.05669466912513599,
|
| 610 |
+
"clip_ratio/region_mean": 0.07041154580656439,
|
| 611 |
+
"epoch": 0.3545706371191136,
|
| 612 |
+
"grad_norm": 0.7437646652660285,
|
| 613 |
+
"kl": 0.3698675720952451,
|
| 614 |
+
"learning_rate": 0.00019120447901834706,
|
| 615 |
+
"loss": -0.013283709064126015,
|
| 616 |
+
"memory(GiB)": 80.03,
|
| 617 |
+
"step": 32,
|
| 618 |
+
"train_speed(iter/s)": 0.002345
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"clip_ratio/high_max": 0.0013764221512246877,
|
| 622 |
+
"clip_ratio/high_mean": 0.0013764221512246877,
|
| 623 |
+
"clip_ratio/low_mean": 0.0013627433363581076,
|
| 624 |
+
"clip_ratio/low_min": 0.0013627433363581076,
|
| 625 |
+
"clip_ratio/region_mean": 0.0027391654875827953,
|
| 626 |
+
"completions/clipped_ratio": 0.0625,
|
| 627 |
+
"completions/max_length": 101.0,
|
| 628 |
+
"completions/mean_length": 24.609375,
|
| 629 |
+
"completions/min_length": 4.0,
|
| 630 |
+
"epoch": 0.3656509695290859,
|
| 631 |
+
"grad_norm": 0.8795930368140337,
|
| 632 |
+
"kl": 0.22658177139237523,
|
| 633 |
+
"learning_rate": 0.00019043571606975777,
|
| 634 |
+
"loss": 0.0019196830689907074,
|
| 635 |
+
"memory(GiB)": 80.03,
|
| 636 |
+
"reward": 0.4140625,
|
| 637 |
+
"reward_std": 0.8300054669380188,
|
| 638 |
+
"rewards/Response_no_think_reward/mean": 0.4140625,
|
| 639 |
+
"rewards/Response_no_think_reward/std": 1.258216142654419,
|
| 640 |
+
"step": 33,
|
| 641 |
+
"train_speed(iter/s)": 0.002323
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"clip_ratio/high_max": 0.016442283987998962,
|
| 645 |
+
"clip_ratio/high_mean": 0.016442283987998962,
|
| 646 |
+
"clip_ratio/low_mean": 0.05633212422253564,
|
| 647 |
+
"clip_ratio/low_min": 0.05633212422253564,
|
| 648 |
+
"clip_ratio/region_mean": 0.07277440826874226,
|
| 649 |
+
"epoch": 0.3767313019390582,
|
| 650 |
+
"grad_norm": 0.5188690402432328,
|
| 651 |
+
"kl": 0.23683911142870784,
|
| 652 |
+
"learning_rate": 0.00018963642955370201,
|
| 653 |
+
"loss": -0.016352392733097076,
|
| 654 |
+
"memory(GiB)": 80.03,
|
| 655 |
+
"step": 34,
|
| 656 |
+
"train_speed(iter/s)": 0.002362
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"clip_ratio/high_max": 0.0025397460121894255,
|
| 660 |
+
"clip_ratio/high_mean": 0.0025397460121894255,
|
| 661 |
+
"clip_ratio/low_mean": 0.00378477135382127,
|
| 662 |
+
"clip_ratio/low_min": 0.00378477135382127,
|
| 663 |
+
"clip_ratio/region_mean": 0.006324517366010696,
|
| 664 |
+
"completions/clipped_ratio": 0.015625,
|
| 665 |
+
"completions/max_length": 101.0,
|
| 666 |
+
"completions/mean_length": 23.375,
|
| 667 |
+
"completions/min_length": 3.0,
|
| 668 |
+
"epoch": 0.3878116343490305,
|
| 669 |
+
"grad_norm": 0.7751350107093148,
|
| 670 |
+
"kl": 0.21053988160565495,
|
| 671 |
+
"learning_rate": 0.00018880688924275378,
|
| 672 |
+
"loss": 0.001979985274374485,
|
| 673 |
+
"memory(GiB)": 80.03,
|
| 674 |
+
"reward": 0.25,
|
| 675 |
+
"reward_std": 0.8051205277442932,
|
| 676 |
+
"rewards/Response_no_think_reward/mean": 0.25,
|
| 677 |
+
"rewards/Response_no_think_reward/std": 1.3101662397384644,
|
| 678 |
+
"step": 35,
|
| 679 |
+
"train_speed(iter/s)": 0.00234
|
| 680 |
+
},
|
| 681 |
+
{
|
| 682 |
+
"clip_ratio/high_max": 0.02383261898648925,
|
| 683 |
+
"clip_ratio/high_mean": 0.02383261898648925,
|
| 684 |
+
"clip_ratio/low_mean": 0.02296329999808222,
|
| 685 |
+
"clip_ratio/low_min": 0.02296329999808222,
|
| 686 |
+
"clip_ratio/region_mean": 0.046795917907729745,
|
| 687 |
+
"epoch": 0.3988919667590028,
|
| 688 |
+
"grad_norm": 0.6919043847160915,
|
| 689 |
+
"kl": 0.2092050458304584,
|
| 690 |
+
"learning_rate": 0.0001879473751206489,
|
| 691 |
+
"loss": -0.016633104532957077,
|
| 692 |
+
"memory(GiB)": 80.03,
|
| 693 |
+
"step": 36,
|
| 694 |
+
"train_speed(iter/s)": 0.002376
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
"clip_ratio/high_max": 0.0,
|
| 698 |
+
"clip_ratio/high_mean": 0.0,
|
| 699 |
+
"clip_ratio/low_mean": 0.0003396739193703979,
|
| 700 |
+
"clip_ratio/low_min": 0.0003396739193703979,
|
| 701 |
+
"clip_ratio/region_mean": 0.0003396739193703979,
|
| 702 |
+
"completions/clipped_ratio": 0.0078125,
|
| 703 |
+
"completions/max_length": 101.0,
|
| 704 |
+
"completions/mean_length": 16.8828125,
|
| 705 |
+
"completions/min_length": 3.0,
|
| 706 |
+
"epoch": 0.4099722991689751,
|
| 707 |
+
"grad_norm": 1.0548464052821087,
|
| 708 |
+
"kl": 0.20961805986007676,
|
| 709 |
+
"learning_rate": 0.00018705817728778624,
|
| 710 |
+
"loss": 0.003657686058431864,
|
| 711 |
+
"memory(GiB)": 80.03,
|
| 712 |
+
"reward": 0.953125,
|
| 713 |
+
"reward_std": 0.6187193393707275,
|
| 714 |
+
"rewards/Response_no_think_reward/mean": 0.953125,
|
| 715 |
+
"rewards/Response_no_think_reward/std": 1.1961840391159058,
|
| 716 |
+
"step": 37,
|
| 717 |
+
"train_speed(iter/s)": 0.002358
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"clip_ratio/high_max": 0.02387295541120693,
|
| 721 |
+
"clip_ratio/high_mean": 0.02387295541120693,
|
| 722 |
+
"clip_ratio/low_mean": 0.030282753868959844,
|
| 723 |
+
"clip_ratio/low_min": 0.030282753868959844,
|
| 724 |
+
"clip_ratio/region_mean": 0.05415570852346718,
|
| 725 |
+
"epoch": 0.42105263157894735,
|
| 726 |
+
"grad_norm": 0.49371052672691246,
|
| 727 |
+
"kl": 0.24616074899677187,
|
| 728 |
+
"learning_rate": 0.00018613959586331362,
|
| 729 |
+
"loss": -0.012075964361429214,
|
| 730 |
+
"memory(GiB)": 80.03,
|
| 731 |
+
"step": 38,
|
| 732 |
+
"train_speed(iter/s)": 0.002393
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"clip_ratio/high_max": 0.0013275333330966532,
|
| 736 |
+
"clip_ratio/high_mean": 0.0013275333330966532,
|
| 737 |
+
"clip_ratio/low_mean": 0.0013904034567531198,
|
| 738 |
+
"clip_ratio/low_min": 0.0013904034567531198,
|
| 739 |
+
"clip_ratio/region_mean": 0.002717936789849773,
|
| 740 |
+
"completions/clipped_ratio": 0.03125,
|
| 741 |
+
"completions/max_length": 101.0,
|
| 742 |
+
"completions/mean_length": 21.6015625,
|
| 743 |
+
"completions/min_length": 4.0,
|
| 744 |
+
"epoch": 0.43213296398891965,
|
| 745 |
+
"grad_norm": 0.972842541028031,
|
| 746 |
+
"kl": 0.16011726018041372,
|
| 747 |
+
"learning_rate": 0.00018519194088383273,
|
| 748 |
+
"loss": 0.001440724590793252,
|
| 749 |
+
"memory(GiB)": 80.03,
|
| 750 |
+
"reward": 0.625,
|
| 751 |
+
"reward_std": 0.6670520305633545,
|
| 752 |
+
"rewards/Response_no_think_reward/mean": 0.625,
|
| 753 |
+
"rewards/Response_no_think_reward/std": 1.3457428216934204,
|
| 754 |
+
"step": 39,
|
| 755 |
+
"train_speed(iter/s)": 0.002371
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"clip_ratio/high_max": 0.03226059180451557,
|
| 759 |
+
"clip_ratio/high_mean": 0.03226059180451557,
|
| 760 |
+
"clip_ratio/low_mean": 0.03314233338460326,
|
| 761 |
+
"clip_ratio/low_min": 0.03314233338460326,
|
| 762 |
+
"clip_ratio/region_mean": 0.0654029252473265,
|
| 763 |
+
"epoch": 0.44321329639889195,
|
| 764 |
+
"grad_norm": 0.5253408277855696,
|
| 765 |
+
"kl": 0.18682625680230558,
|
| 766 |
+
"learning_rate": 0.00018421553219875658,
|
| 767 |
+
"loss": -0.013099671341478825,
|
| 768 |
+
"memory(GiB)": 80.03,
|
| 769 |
+
"step": 40,
|
| 770 |
+
"train_speed(iter/s)": 0.002404
|
| 771 |
+
},
|
| 772 |
+
{
|
| 773 |
+
"clip_ratio/high_max": 0.00029620854184031487,
|
| 774 |
+
"clip_ratio/high_mean": 0.00029620854184031487,
|
| 775 |
+
"clip_ratio/low_mean": 0.0009311849280493334,
|
| 776 |
+
"clip_ratio/low_min": 0.0009311849280493334,
|
| 777 |
+
"clip_ratio/region_mean": 0.0012273934698896483,
|
| 778 |
+
"completions/clipped_ratio": 0.0,
|
| 779 |
+
"completions/max_length": 88.0,
|
| 780 |
+
"completions/mean_length": 17.5625,
|
| 781 |
+
"completions/min_length": 5.0,
|
| 782 |
+
"epoch": 0.45429362880886426,
|
| 783 |
+
"grad_norm": 1.2919745010410586,
|
| 784 |
+
"kl": 0.2366366102360189,
|
| 785 |
+
"learning_rate": 0.00018321069936235503,
|
| 786 |
+
"loss": 0.001985038397833705,
|
| 787 |
+
"memory(GiB)": 80.03,
|
| 788 |
+
"reward": 0.859375,
|
| 789 |
+
"reward_std": 0.6107450723648071,
|
| 790 |
+
"rewards/Response_no_think_reward/mean": 0.859375,
|
| 791 |
+
"rewards/Response_no_think_reward/std": 1.4071491956710815,
|
| 792 |
+
"step": 41,
|
| 793 |
+
"train_speed(iter/s)": 0.002379
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"clip_ratio/high_max": 0.01783788768807426,
|
| 797 |
+
"clip_ratio/high_mean": 0.01783788768807426,
|
| 798 |
+
"clip_ratio/low_mean": 0.044289187353570014,
|
| 799 |
+
"clip_ratio/low_min": 0.044289187353570014,
|
| 800 |
+
"clip_ratio/region_mean": 0.06212707597296685,
|
| 801 |
+
"epoch": 0.46537396121883656,
|
| 802 |
+
"grad_norm": 1.7719321167278534,
|
| 803 |
+
"kl": 0.6719344789162278,
|
| 804 |
+
"learning_rate": 0.0001821777815225245,
|
| 805 |
+
"loss": -0.008167732506990433,
|
| 806 |
+
"memory(GiB)": 80.03,
|
| 807 |
+
"step": 42,
|
| 808 |
+
"train_speed(iter/s)": 0.00241
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"clip_ratio/high_max": 0.00036231885314919055,
|
| 812 |
+
"clip_ratio/high_mean": 0.00036231885314919055,
|
| 813 |
+
"clip_ratio/low_mean": 0.002238474931800738,
|
| 814 |
+
"clip_ratio/low_min": 0.002238474931800738,
|
| 815 |
+
"clip_ratio/region_mean": 0.0026007937849499285,
|
| 816 |
+
"completions/clipped_ratio": 0.0,
|
| 817 |
+
"completions/max_length": 97.0,
|
| 818 |
+
"completions/mean_length": 15.875,
|
| 819 |
+
"completions/min_length": 6.0,
|
| 820 |
+
"epoch": 0.47645429362880887,
|
| 821 |
+
"grad_norm": 1.1682612472028975,
|
| 822 |
+
"kl": 0.2416230053640902,
|
| 823 |
+
"learning_rate": 0.00018111712730632022,
|
| 824 |
+
"loss": 0.0018095734994858503,
|
| 825 |
+
"memory(GiB)": 80.03,
|
| 826 |
+
"reward": 0.0859375,
|
| 827 |
+
"reward_std": 0.39637458324432373,
|
| 828 |
+
"rewards/Response_no_think_reward/mean": 0.0859375,
|
| 829 |
+
"rewards/Response_no_think_reward/std": 0.9962713122367859,
|
| 830 |
+
"step": 43,
|
| 831 |
+
"train_speed(iter/s)": 0.002397
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"clip_ratio/high_max": 0.007687599485507235,
|
| 835 |
+
"clip_ratio/high_mean": 0.007687599485507235,
|
| 836 |
+
"clip_ratio/low_mean": 0.042899149731965736,
|
| 837 |
+
"clip_ratio/low_min": 0.042899149731965736,
|
| 838 |
+
"clip_ratio/region_mean": 0.05058674863539636,
|
| 839 |
+
"epoch": 0.48753462603878117,
|
| 840 |
+
"grad_norm": 17.524605028595435,
|
| 841 |
+
"kl": 15.062655651359819,
|
| 842 |
+
"learning_rate": 0.00018002909470228842,
|
| 843 |
+
"loss": 0.09481670707464218,
|
| 844 |
+
"memory(GiB)": 80.03,
|
| 845 |
+
"step": 44,
|
| 846 |
+
"train_speed(iter/s)": 0.002427
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"clip_ratio/high_max": 0.0012922932510264218,
|
| 850 |
+
"clip_ratio/high_mean": 0.0012922932510264218,
|
| 851 |
+
"clip_ratio/low_mean": 0.002554390055593103,
|
| 852 |
+
"clip_ratio/low_min": 0.002554390055593103,
|
| 853 |
+
"clip_ratio/region_mean": 0.003846683306619525,
|
| 854 |
+
"completions/clipped_ratio": 0.0,
|
| 855 |
+
"completions/max_length": 32.0,
|
| 856 |
+
"completions/mean_length": 8.546875,
|
| 857 |
+
"completions/min_length": 4.0,
|
| 858 |
+
"epoch": 0.4986149584487535,
|
| 859 |
+
"grad_norm": 1.474323820635427,
|
| 860 |
+
"kl": 0.7951482257340103,
|
| 861 |
+
"learning_rate": 0.00017891405093963938,
|
| 862 |
+
"loss": 0.008442065678536892,
|
| 863 |
+
"memory(GiB)": 80.03,
|
| 864 |
+
"reward": 0.171875,
|
| 865 |
+
"reward_std": 0.540536105632782,
|
| 866 |
+
"rewards/Response_no_think_reward/mean": 0.171875,
|
| 867 |
+
"rewards/Response_no_think_reward/std": 0.9648089408874512,
|
| 868 |
+
"step": 45,
|
| 869 |
+
"train_speed(iter/s)": 0.002416
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"clip_ratio/high_max": 0.01782548933988437,
|
| 873 |
+
"clip_ratio/high_mean": 0.01782548933988437,
|
| 874 |
+
"clip_ratio/low_mean": 0.04517949424916878,
|
| 875 |
+
"clip_ratio/low_min": 0.04517949424916878,
|
| 876 |
+
"clip_ratio/region_mean": 0.06300498393829912,
|
| 877 |
+
"epoch": 0.5096952908587258,
|
| 878 |
+
"grad_norm": 23.18489815639863,
|
| 879 |
+
"kl": 20.399557466851547,
|
| 880 |
+
"learning_rate": 0.0001777723723643014,
|
| 881 |
+
"loss": 0.15100935101509094,
|
| 882 |
+
"memory(GiB)": 80.03,
|
| 883 |
+
"step": 46,
|
| 884 |
+
"train_speed(iter/s)": 0.002445
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"clip_ratio/high_max": 0.0001338329748250544,
|
| 888 |
+
"clip_ratio/high_mean": 0.0001338329748250544,
|
| 889 |
+
"clip_ratio/low_mean": 0.0020056332577951252,
|
| 890 |
+
"clip_ratio/low_min": 0.0020056332577951252,
|
| 891 |
+
"clip_ratio/region_mean": 0.0021394662326201797,
|
| 892 |
+
"completions/clipped_ratio": 0.015625,
|
| 893 |
+
"completions/max_length": 101.0,
|
| 894 |
+
"completions/mean_length": 11.8203125,
|
| 895 |
+
"completions/min_length": 4.0,
|
| 896 |
+
"epoch": 0.5207756232686981,
|
| 897 |
+
"grad_norm": 1.9101810703272282,
|
| 898 |
+
"kl": 0.8325624349527061,
|
| 899 |
+
"learning_rate": 0.0001766044443118978,
|
| 900 |
+
"loss": 0.01317879930138588,
|
| 901 |
+
"memory(GiB)": 80.03,
|
| 902 |
+
"reward": 0.21875,
|
| 903 |
+
"reward_std": 0.5618736743927002,
|
| 904 |
+
"rewards/Response_no_think_reward/mean": 0.21875,
|
| 905 |
+
"rewards/Response_no_think_reward/std": 0.9301384091377258,
|
| 906 |
+
"step": 47,
|
| 907 |
+
"train_speed(iter/s)": 0.002435
|
| 908 |
+
},
|
| 909 |
+
{
|
| 910 |
+
"clip_ratio/high_max": 0.01570215745596215,
|
| 911 |
+
"clip_ratio/high_mean": 0.01570215745596215,
|
| 912 |
+
"clip_ratio/low_mean": 0.01864259922876954,
|
| 913 |
+
"clip_ratio/low_min": 0.01864259922876954,
|
| 914 |
+
"clip_ratio/region_mean": 0.03434475651010871,
|
| 915 |
+
"epoch": 0.5318559556786704,
|
| 916 |
+
"grad_norm": 5.211743253538076,
|
| 917 |
+
"kl": 0.4838231955654919,
|
| 918 |
+
"learning_rate": 0.00017541066097768963,
|
| 919 |
+
"loss": 0.03044246882200241,
|
| 920 |
+
"memory(GiB)": 80.03,
|
| 921 |
+
"step": 48,
|
| 922 |
+
"train_speed(iter/s)": 0.002462
|
| 923 |
+
},
|
| 924 |
+
{
|
| 925 |
+
"clip_ratio/high_max": 0.0,
|
| 926 |
+
"clip_ratio/high_mean": 0.0,
|
| 927 |
+
"clip_ratio/low_mean": 0.002905012297560461,
|
| 928 |
+
"clip_ratio/low_min": 0.002905012297560461,
|
| 929 |
+
"clip_ratio/region_mean": 0.002905012297560461,
|
| 930 |
+
"completions/clipped_ratio": 0.0390625,
|
| 931 |
+
"completions/max_length": 101.0,
|
| 932 |
+
"completions/mean_length": 17.7265625,
|
| 933 |
+
"completions/min_length": 4.0,
|
| 934 |
+
"epoch": 0.5429362880886427,
|
| 935 |
+
"grad_norm": 0.9459153229739418,
|
| 936 |
+
"kl": 0.33009129445417784,
|
| 937 |
+
"learning_rate": 0.00017419142528352817,
|
| 938 |
+
"loss": 0.004404420033097267,
|
| 939 |
+
"memory(GiB)": 80.03,
|
| 940 |
+
"reward": -0.0859375,
|
| 941 |
+
"reward_std": 0.547653079032898,
|
| 942 |
+
"rewards/Response_no_think_reward/mean": -0.0859375,
|
| 943 |
+
"rewards/Response_no_think_reward/std": 0.9138250946998596,
|
| 944 |
+
"step": 49,
|
| 945 |
+
"train_speed(iter/s)": 0.002452
|
| 946 |
+
},
|
| 947 |
+
{
|
| 948 |
+
"clip_ratio/high_max": 0.00849696435034275,
|
| 949 |
+
"clip_ratio/high_mean": 0.00849696435034275,
|
| 950 |
+
"clip_ratio/low_mean": 0.059440263896249235,
|
| 951 |
+
"clip_ratio/low_min": 0.059440263896249235,
|
| 952 |
+
"clip_ratio/region_mean": 0.06793722766451538,
|
| 953 |
+
"epoch": 0.554016620498615,
|
| 954 |
+
"grad_norm": 1.5192885268898135,
|
| 955 |
+
"kl": 0.5284001431518845,
|
| 956 |
+
"learning_rate": 0.0001729471487418621,
|
| 957 |
+
"loss": -0.012666964903473854,
|
| 958 |
+
"memory(GiB)": 80.03,
|
| 959 |
+
"step": 50,
|
| 960 |
+
"train_speed(iter/s)": 0.002478
|
| 961 |
+
},
|
| 962 |
+
{
|
| 963 |
+
"clip_ratio/high_max": 0.0,
|
| 964 |
+
"clip_ratio/high_mean": 0.0,
|
| 965 |
+
"clip_ratio/low_mean": 0.0012755101779475808,
|
| 966 |
+
"clip_ratio/low_min": 0.0012755101779475808,
|
| 967 |
+
"clip_ratio/region_mean": 0.0012755101779475808,
|
| 968 |
+
"completions/clipped_ratio": 0.0078125,
|
| 969 |
+
"completions/max_length": 101.0,
|
| 970 |
+
"completions/mean_length": 12.640625,
|
| 971 |
+
"completions/min_length": 3.0,
|
| 972 |
+
"epoch": 0.5650969529085873,
|
| 973 |
+
"grad_norm": 1.7932835802898264,
|
| 974 |
+
"kl": 0.582448753528297,
|
| 975 |
+
"learning_rate": 0.00017167825131684513,
|
| 976 |
+
"loss": 0.008432275615632534,
|
| 977 |
+
"memory(GiB)": 80.03,
|
| 978 |
+
"reward": 0.15625,
|
| 979 |
+
"reward_std": 0.7085258960723877,
|
| 980 |
+
"rewards/Response_no_think_reward/mean": 0.15625,
|
| 981 |
+
"rewards/Response_no_think_reward/std": 1.0228685140609741,
|
| 982 |
+
"step": 51,
|
| 983 |
+
"train_speed(iter/s)": 0.002465
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"clip_ratio/high_max": 0.006597792147658765,
|
| 987 |
+
"clip_ratio/high_mean": 0.006597792147658765,
|
| 988 |
+
"clip_ratio/low_mean": 0.049801092012785375,
|
| 989 |
+
"clip_ratio/low_min": 0.049801092012785375,
|
| 990 |
+
"clip_ratio/region_mean": 0.05639888462610543,
|
| 991 |
+
"epoch": 0.5761772853185596,
|
| 992 |
+
"grad_norm": 0.8422307837338449,
|
| 993 |
+
"kl": 0.7035694038495421,
|
| 994 |
+
"learning_rate": 0.00017038516128259115,
|
| 995 |
+
"loss": -0.011608053930103779,
|
| 996 |
+
"memory(GiB)": 80.03,
|
| 997 |
+
"step": 52,
|
| 998 |
+
"train_speed(iter/s)": 0.00249
|
| 999 |
+
},
|
| 1000 |
+
{
|
| 1001 |
+
"clip_ratio/high_max": 0.0,
|
| 1002 |
+
"clip_ratio/high_mean": 0.0,
|
| 1003 |
+
"clip_ratio/low_mean": 0.0005558648990700021,
|
| 1004 |
+
"clip_ratio/low_min": 0.0005558648990700021,
|
| 1005 |
+
"clip_ratio/region_mean": 0.0005558648990700021,
|
| 1006 |
+
"completions/clipped_ratio": 0.015625,
|
| 1007 |
+
"completions/max_length": 101.0,
|
| 1008 |
+
"completions/mean_length": 12.5,
|
| 1009 |
+
"completions/min_length": 4.0,
|
| 1010 |
+
"epoch": 0.5872576177285319,
|
| 1011 |
+
"grad_norm": 1.0211580850819229,
|
| 1012 |
+
"kl": 0.7616109761074767,
|
| 1013 |
+
"learning_rate": 0.00016906831507862443,
|
| 1014 |
+
"loss": 0.005634145811200142,
|
| 1015 |
+
"memory(GiB)": 80.03,
|
| 1016 |
+
"reward": 0.1640625,
|
| 1017 |
+
"reward_std": 0.70704185962677,
|
| 1018 |
+
"rewards/Response_no_think_reward/mean": 0.1640625,
|
| 1019 |
+
"rewards/Response_no_think_reward/std": 0.9702450633049011,
|
| 1020 |
+
"step": 53,
|
| 1021 |
+
"train_speed(iter/s)": 0.002477
|
| 1022 |
+
},
|
| 1023 |
+
{
|
| 1024 |
+
"clip_ratio/high_max": 0.02744574609096162,
|
| 1025 |
+
"clip_ratio/high_mean": 0.02744574609096162,
|
| 1026 |
+
"clip_ratio/low_mean": 0.04221567645436153,
|
| 1027 |
+
"clip_ratio/low_min": 0.04221567645436153,
|
| 1028 |
+
"clip_ratio/region_mean": 0.06966142146848142,
|
| 1029 |
+
"epoch": 0.5983379501385041,
|
| 1030 |
+
"grad_norm": 0.5808426269905869,
|
| 1031 |
+
"kl": 0.801087921798171,
|
| 1032 |
+
"learning_rate": 0.00016772815716257412,
|
| 1033 |
+
"loss": -0.01366308145225048,
|
| 1034 |
+
"memory(GiB)": 80.03,
|
| 1035 |
+
"step": 54,
|
| 1036 |
+
"train_speed(iter/s)": 0.002502
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"clip_ratio/high_max": 0.0003881987649947405,
|
| 1040 |
+
"clip_ratio/high_mean": 0.0003881987649947405,
|
| 1041 |
+
"clip_ratio/low_mean": 0.001932911021867767,
|
| 1042 |
+
"clip_ratio/low_min": 0.001932911021867767,
|
| 1043 |
+
"clip_ratio/region_mean": 0.0023211097868625075,
|
| 1044 |
+
"completions/clipped_ratio": 0.0390625,
|
| 1045 |
+
"completions/max_length": 101.0,
|
| 1046 |
+
"completions/mean_length": 18.9921875,
|
| 1047 |
+
"completions/min_length": 6.0,
|
| 1048 |
+
"epoch": 0.6094182825484764,
|
| 1049 |
+
"grad_norm": 1.1265096543789663,
|
| 1050 |
+
"kl": 0.5226203491911292,
|
| 1051 |
+
"learning_rate": 0.00016636513986016213,
|
| 1052 |
+
"loss": 0.007094700820744038,
|
| 1053 |
+
"memory(GiB)": 80.03,
|
| 1054 |
+
"reward": 0.1875,
|
| 1055 |
+
"reward_std": 0.7182328104972839,
|
| 1056 |
+
"rewards/Response_no_think_reward/mean": 0.1875,
|
| 1057 |
+
"rewards/Response_no_think_reward/std": 1.0480577945709229,
|
| 1058 |
+
"step": 55,
|
| 1059 |
+
"train_speed(iter/s)": 0.00249
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"clip_ratio/high_max": 0.011908911721548066,
|
| 1063 |
+
"clip_ratio/high_mean": 0.011908911721548066,
|
| 1064 |
+
"clip_ratio/low_mean": 0.05817525731981732,
|
| 1065 |
+
"clip_ratio/low_min": 0.05817525731981732,
|
| 1066 |
+
"clip_ratio/region_mean": 0.07008416869211942,
|
| 1067 |
+
"epoch": 0.6204986149584487,
|
| 1068 |
+
"grad_norm": 0.5560920258062684,
|
| 1069 |
+
"kl": 0.5747523186728358,
|
| 1070 |
+
"learning_rate": 0.000164979723212536,
|
| 1071 |
+
"loss": -0.014954826794564724,
|
| 1072 |
+
"memory(GiB)": 80.03,
|
| 1073 |
+
"step": 56,
|
| 1074 |
+
"train_speed(iter/s)": 0.002513
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"clip_ratio/high_max": 0.002692167239729315,
|
| 1078 |
+
"clip_ratio/high_mean": 0.002692167239729315,
|
| 1079 |
+
"clip_ratio/low_mean": 0.0006428283377317712,
|
| 1080 |
+
"clip_ratio/low_min": 0.0006428283377317712,
|
| 1081 |
+
"clip_ratio/region_mean": 0.003334995577461086,
|
| 1082 |
+
"completions/clipped_ratio": 0.0390625,
|
| 1083 |
+
"completions/max_length": 101.0,
|
| 1084 |
+
"completions/mean_length": 16.53125,
|
| 1085 |
+
"completions/min_length": 6.0,
|
| 1086 |
+
"epoch": 0.631578947368421,
|
| 1087 |
+
"grad_norm": 1.5596809103877662,
|
| 1088 |
+
"kl": 0.6855434570461512,
|
| 1089 |
+
"learning_rate": 0.00016357237482099684,
|
| 1090 |
+
"loss": 0.0062956069596111774,
|
| 1091 |
+
"memory(GiB)": 80.03,
|
| 1092 |
+
"reward": 0.53125,
|
| 1093 |
+
"reward_std": 0.5907745361328125,
|
| 1094 |
+
"rewards/Response_no_think_reward/mean": 0.53125,
|
| 1095 |
+
"rewards/Response_no_think_reward/std": 1.0790598392486572,
|
| 1096 |
+
"step": 57,
|
| 1097 |
+
"train_speed(iter/s)": 0.002501
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"clip_ratio/high_max": 0.0316368987550959,
|
| 1101 |
+
"clip_ratio/high_mean": 0.0316368987550959,
|
| 1102 |
+
"clip_ratio/low_mean": 0.015557856269879267,
|
| 1103 |
+
"clip_ratio/low_min": 0.015557856269879267,
|
| 1104 |
+
"clip_ratio/region_mean": 0.04719475514139049,
|
| 1105 |
+
"epoch": 0.6426592797783933,
|
| 1106 |
+
"grad_norm": 0.950735686233936,
|
| 1107 |
+
"kl": 0.5240823943167925,
|
| 1108 |
+
"learning_rate": 0.00016214356968917648,
|
| 1109 |
+
"loss": -0.015531505458056927,
|
| 1110 |
+
"memory(GiB)": 80.03,
|
| 1111 |
+
"step": 58,
|
| 1112 |
+
"train_speed(iter/s)": 0.002524
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"clip_ratio/high_max": 0.00011467889999039471,
|
| 1116 |
+
"clip_ratio/high_mean": 0.00011467889999039471,
|
| 1117 |
+
"clip_ratio/low_mean": 0.0004376750875962898,
|
| 1118 |
+
"clip_ratio/low_min": 0.0004376750875962898,
|
| 1119 |
+
"clip_ratio/region_mean": 0.0005523539875866845,
|
| 1120 |
+
"completions/clipped_ratio": 0.078125,
|
| 1121 |
+
"completions/max_length": 101.0,
|
| 1122 |
+
"completions/mean_length": 23.9140625,
|
| 1123 |
+
"completions/min_length": 6.0,
|
| 1124 |
+
"epoch": 0.6537396121883656,
|
| 1125 |
+
"grad_norm": 1.3162721482480142,
|
| 1126 |
+
"kl": 0.5214177745606321,
|
| 1127 |
+
"learning_rate": 0.00016069379006271566,
|
| 1128 |
+
"loss": 0.004921327345073223,
|
| 1129 |
+
"memory(GiB)": 80.03,
|
| 1130 |
+
"reward": -0.0625,
|
| 1131 |
+
"reward_std": 0.6303451657295227,
|
| 1132 |
+
"rewards/Response_no_think_reward/mean": -0.0625,
|
| 1133 |
+
"rewards/Response_no_think_reward/std": 1.0019665956497192,
|
| 1134 |
+
"step": 59,
|
| 1135 |
+
"train_speed(iter/s)": 0.002511
|
| 1136 |
+
},
|
| 1137 |
+
{
|
| 1138 |
+
"clip_ratio/high_max": 0.00810479320352897,
|
| 1139 |
+
"clip_ratio/high_mean": 0.00810479320352897,
|
| 1140 |
+
"clip_ratio/low_mean": 0.05729365168372169,
|
| 1141 |
+
"clip_ratio/low_min": 0.05729365168372169,
|
| 1142 |
+
"clip_ratio/region_mean": 0.06539844395592809,
|
| 1143 |
+
"epoch": 0.6648199445983379,
|
| 1144 |
+
"grad_norm": 0.5733548474442324,
|
| 1145 |
+
"kl": 0.6363338100200053,
|
| 1146 |
+
"learning_rate": 0.00015922352526649803,
|
| 1147 |
+
"loss": -0.021113965660333633,
|
| 1148 |
+
"memory(GiB)": 80.03,
|
| 1149 |
+
"step": 60,
|
| 1150 |
+
"train_speed(iter/s)": 0.002532
|
| 1151 |
+
},
|
| 1152 |
+
{
|
| 1153 |
+
"clip_ratio/high_max": 0.0005996339968987741,
|
| 1154 |
+
"clip_ratio/high_mean": 0.0005996339968987741,
|
| 1155 |
+
"clip_ratio/low_mean": 0.004901745676761493,
|
| 1156 |
+
"clip_ratio/low_min": 0.004901745676761493,
|
| 1157 |
+
"clip_ratio/region_mean": 0.00550137969548814,
|
| 1158 |
+
"completions/clipped_ratio": 0.0859375,
|
| 1159 |
+
"completions/max_length": 101.0,
|
| 1160 |
+
"completions/mean_length": 22.375,
|
| 1161 |
+
"completions/min_length": 2.0,
|
| 1162 |
+
"epoch": 0.6759002770083102,
|
| 1163 |
+
"grad_norm": 1.887884786225259,
|
| 1164 |
+
"kl": 2.07791917472332,
|
| 1165 |
+
"learning_rate": 0.00015773327153949465,
|
| 1166 |
+
"loss": 0.008165515027940273,
|
| 1167 |
+
"memory(GiB)": 80.03,
|
| 1168 |
+
"reward": 0.40625,
|
| 1169 |
+
"reward_std": 0.5973243117332458,
|
| 1170 |
+
"rewards/Response_no_think_reward/mean": 0.40625,
|
| 1171 |
+
"rewards/Response_no_think_reward/std": 1.186787486076355,
|
| 1172 |
+
"step": 61,
|
| 1173 |
+
"train_speed(iter/s)": 0.002506
|
| 1174 |
+
},
|
| 1175 |
+
{
|
| 1176 |
+
"clip_ratio/high_max": 0.017942053091246635,
|
| 1177 |
+
"clip_ratio/high_mean": 0.017942053091246635,
|
| 1178 |
+
"clip_ratio/low_mean": 0.06714771036058664,
|
| 1179 |
+
"clip_ratio/low_min": 0.06714771036058664,
|
| 1180 |
+
"clip_ratio/region_mean": 0.08508976292796433,
|
| 1181 |
+
"epoch": 0.6869806094182825,
|
| 1182 |
+
"grad_norm": 11.312330346219078,
|
| 1183 |
+
"kl": 0.5888316835043952,
|
| 1184 |
+
"learning_rate": 0.00015622353186727544,
|
| 1185 |
+
"loss": 0.04619387909770012,
|
| 1186 |
+
"memory(GiB)": 80.03,
|
| 1187 |
+
"step": 62,
|
| 1188 |
+
"train_speed(iter/s)": 0.002527
|
| 1189 |
+
},
|
| 1190 |
+
{
|
| 1191 |
+
"clip_ratio/high_max": 0.00020032051543239504,
|
| 1192 |
+
"clip_ratio/high_mean": 0.00020032051543239504,
|
| 1193 |
+
"clip_ratio/low_mean": 0.001202335930429399,
|
| 1194 |
+
"clip_ratio/low_min": 0.001202335930429399,
|
| 1195 |
+
"clip_ratio/region_mean": 0.001402656445861794,
|
| 1196 |
+
"completions/clipped_ratio": 0.0546875,
|
| 1197 |
+
"completions/max_length": 101.0,
|
| 1198 |
+
"completions/mean_length": 18.5859375,
|
| 1199 |
+
"completions/min_length": 2.0,
|
| 1200 |
+
"epoch": 0.6980609418282548,
|
| 1201 |
+
"grad_norm": 1.9804020858052087,
|
| 1202 |
+
"kl": 6.258792589243967,
|
| 1203 |
+
"learning_rate": 0.00015469481581224272,
|
| 1204 |
+
"loss": 0.014128579758107662,
|
| 1205 |
+
"memory(GiB)": 80.03,
|
| 1206 |
+
"reward": 0.78125,
|
| 1207 |
+
"reward_std": 0.5936684608459473,
|
| 1208 |
+
"rewards/Response_no_think_reward/mean": 0.78125,
|
| 1209 |
+
"rewards/Response_no_think_reward/std": 1.235546350479126,
|
| 1210 |
+
"step": 63,
|
| 1211 |
+
"train_speed(iter/s)": 0.002486
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"clip_ratio/high_max": 0.020077986438991502,
|
| 1215 |
+
"clip_ratio/high_mean": 0.020077986438991502,
|
| 1216 |
+
"clip_ratio/low_mean": 0.12254823022522032,
|
| 1217 |
+
"clip_ratio/low_min": 0.12254823022522032,
|
| 1218 |
+
"clip_ratio/region_mean": 0.142626216635108,
|
| 1219 |
+
"epoch": 0.7091412742382271,
|
| 1220 |
+
"grad_norm": 1.5460664241155249,
|
| 1221 |
+
"kl": 5.241092938755173,
|
| 1222 |
+
"learning_rate": 0.0001531476393416456,
|
| 1223 |
+
"loss": -0.0033248686231672764,
|
| 1224 |
+
"memory(GiB)": 80.03,
|
| 1225 |
+
"step": 64,
|
| 1226 |
+
"train_speed(iter/s)": 0.002506
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"clip_ratio/high_max": 0.0006774475477868691,
|
| 1230 |
+
"clip_ratio/high_mean": 0.0006774475477868691,
|
| 1231 |
+
"clip_ratio/low_mean": 0.0009178321633953601,
|
| 1232 |
+
"clip_ratio/low_min": 0.0009178321633953601,
|
| 1233 |
+
"clip_ratio/region_mean": 0.0015952797257341444,
|
| 1234 |
+
"completions/clipped_ratio": 0.0390625,
|
| 1235 |
+
"completions/max_length": 101.0,
|
| 1236 |
+
"completions/mean_length": 15.1484375,
|
| 1237 |
+
"completions/min_length": 6.0,
|
| 1238 |
+
"epoch": 0.7202216066481995,
|
| 1239 |
+
"grad_norm": 1.528383760465238,
|
| 1240 |
+
"kl": 1.888367731589824,
|
| 1241 |
+
"learning_rate": 0.00015158252465343242,
|
| 1242 |
+
"loss": 0.010861902497708797,
|
| 1243 |
+
"memory(GiB)": 80.03,
|
| 1244 |
+
"reward": 0.59375,
|
| 1245 |
+
"reward_std": 0.5738716125488281,
|
| 1246 |
+
"rewards/Response_no_think_reward/mean": 0.59375,
|
| 1247 |
+
"rewards/Response_no_think_reward/std": 1.1462881565093994,
|
| 1248 |
+
"step": 65,
|
| 1249 |
+
"train_speed(iter/s)": 0.002444
|
| 1250 |
+
},
|
| 1251 |
+
{
|
| 1252 |
+
"clip_ratio/high_max": 0.05604529404081404,
|
| 1253 |
+
"clip_ratio/high_mean": 0.05604529404081404,
|
| 1254 |
+
"clip_ratio/low_mean": 0.012011443439405411,
|
| 1255 |
+
"clip_ratio/low_min": 0.012011443439405411,
|
| 1256 |
+
"clip_ratio/region_mean": 0.06805673893541098,
|
| 1257 |
+
"epoch": 0.7313019390581718,
|
| 1258 |
+
"grad_norm": 0.9299498266912626,
|
| 1259 |
+
"kl": 1.0373663480422692,
|
| 1260 |
+
"learning_rate": 0.00015000000000000001,
|
| 1261 |
+
"loss": -0.001186850480735302,
|
| 1262 |
+
"memory(GiB)": 80.03,
|
| 1263 |
+
"step": 66,
|
| 1264 |
+
"train_speed(iter/s)": 0.002463
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"clip_ratio/high_max": 0.006279462773818523,
|
| 1268 |
+
"clip_ratio/high_mean": 0.006279462773818523,
|
| 1269 |
+
"clip_ratio/low_mean": 0.006620214961003512,
|
| 1270 |
+
"clip_ratio/low_min": 0.006620214961003512,
|
| 1271 |
+
"clip_ratio/region_mean": 0.012899677676614374,
|
| 1272 |
+
"completions/clipped_ratio": 0.0546875,
|
| 1273 |
+
"completions/max_length": 101.0,
|
| 1274 |
+
"completions/mean_length": 17.5234375,
|
| 1275 |
+
"completions/min_length": 2.0,
|
| 1276 |
+
"epoch": 0.7423822714681441,
|
| 1277 |
+
"grad_norm": 3.0195311219652377,
|
| 1278 |
+
"kl": 1.4996049946639687,
|
| 1279 |
+
"learning_rate": 0.0001484005995098999,
|
| 1280 |
+
"loss": 0.0103817880153656,
|
| 1281 |
+
"memory(GiB)": 80.03,
|
| 1282 |
+
"reward": 0.2421875,
|
| 1283 |
+
"reward_std": 0.5205168724060059,
|
| 1284 |
+
"rewards/Response_no_think_reward/mean": 0.2421875,
|
| 1285 |
+
"rewards/Response_no_think_reward/std": 1.2595843076705933,
|
| 1286 |
+
"step": 67,
|
| 1287 |
+
"train_speed(iter/s)": 0.002399
|
| 1288 |
+
},
|
| 1289 |
+
{
|
| 1290 |
+
"clip_ratio/high_max": 0.030079254298470914,
|
| 1291 |
+
"clip_ratio/high_mean": 0.030079254298470914,
|
| 1292 |
+
"clip_ratio/low_mean": 0.10815927106887102,
|
| 1293 |
+
"clip_ratio/low_min": 0.10815927106887102,
|
| 1294 |
+
"clip_ratio/region_mean": 0.1382385252509266,
|
| 1295 |
+
"epoch": 0.7534626038781164,
|
| 1296 |
+
"grad_norm": 7.7164901906176375,
|
| 1297 |
+
"kl": 4.463950714329258,
|
| 1298 |
+
"learning_rate": 0.0001467848630075608,
|
| 1299 |
+
"loss": 0.0073772999458014965,
|
| 1300 |
+
"memory(GiB)": 80.03,
|
| 1301 |
+
"step": 68,
|
| 1302 |
+
"train_speed(iter/s)": 0.002418
|
| 1303 |
+
},
|
| 1304 |
+
{
|
| 1305 |
+
"clip_ratio/high_max": 0.0009369817780680023,
|
| 1306 |
+
"clip_ratio/high_mean": 0.0009369817780680023,
|
| 1307 |
+
"clip_ratio/low_mean": 0.006284750299528241,
|
| 1308 |
+
"clip_ratio/low_min": 0.006284750299528241,
|
| 1309 |
+
"clip_ratio/region_mean": 0.007221732055768371,
|
| 1310 |
+
"completions/clipped_ratio": 0.1484375,
|
| 1311 |
+
"completions/max_length": 101.0,
|
| 1312 |
+
"completions/mean_length": 26.9296875,
|
| 1313 |
+
"completions/min_length": 3.0,
|
| 1314 |
+
"epoch": 0.7645429362880887,
|
| 1315 |
+
"grad_norm": 2.034534758963297,
|
| 1316 |
+
"kl": 2.321827916195616,
|
| 1317 |
+
"learning_rate": 0.00014515333583108896,
|
| 1318 |
+
"loss": 0.01671188324689865,
|
| 1319 |
+
"memory(GiB)": 80.03,
|
| 1320 |
+
"reward": 0.5390625,
|
| 1321 |
+
"reward_std": 0.4874863028526306,
|
| 1322 |
+
"rewards/Response_no_think_reward/mean": 0.5390625,
|
| 1323 |
+
"rewards/Response_no_think_reward/std": 1.3095791339874268,
|
| 1324 |
+
"step": 69,
|
| 1325 |
+
"train_speed(iter/s)": 0.002373
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"clip_ratio/high_max": 0.012032406637445092,
|
| 1329 |
+
"clip_ratio/high_mean": 0.012032406637445092,
|
| 1330 |
+
"clip_ratio/low_mean": 0.11593639780767262,
|
| 1331 |
+
"clip_ratio/low_min": 0.11593639780767262,
|
| 1332 |
+
"clip_ratio/region_mean": 0.1279688044451177,
|
| 1333 |
+
"epoch": 0.775623268698061,
|
| 1334 |
+
"grad_norm": 1.497770517232466,
|
| 1335 |
+
"kl": 3.323778461664915,
|
| 1336 |
+
"learning_rate": 0.00014350656864820733,
|
| 1337 |
+
"loss": 0.0011269270908087492,
|
| 1338 |
+
"memory(GiB)": 80.03,
|
| 1339 |
+
"step": 70,
|
| 1340 |
+
"train_speed(iter/s)": 0.002392
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"clip_ratio/high_max": 0.000255623715929687,
|
| 1344 |
+
"clip_ratio/high_mean": 0.000255623715929687,
|
| 1345 |
+
"clip_ratio/low_mean": 0.005516766890650615,
|
| 1346 |
+
"clip_ratio/low_min": 0.005516766890650615,
|
| 1347 |
+
"clip_ratio/region_mean": 0.005772390664787963,
|
| 1348 |
+
"completions/clipped_ratio": 0.09375,
|
| 1349 |
+
"completions/max_length": 101.0,
|
| 1350 |
+
"completions/mean_length": 17.25,
|
| 1351 |
+
"completions/min_length": 2.0,
|
| 1352 |
+
"epoch": 0.7867036011080333,
|
| 1353 |
+
"grad_norm": 2.3636642819394384,
|
| 1354 |
+
"kl": 2.6716066980734468,
|
| 1355 |
+
"learning_rate": 0.00014184511727039612,
|
| 1356 |
+
"loss": 0.020374953746795654,
|
| 1357 |
+
"memory(GiB)": 80.03,
|
| 1358 |
+
"reward": 0.34375,
|
| 1359 |
+
"reward_std": 0.5405020713806152,
|
| 1360 |
+
"rewards/Response_no_think_reward/mean": 0.34375,
|
| 1361 |
+
"rewards/Response_no_think_reward/std": 1.2323558330535889,
|
| 1362 |
+
"step": 71,
|
| 1363 |
+
"train_speed(iter/s)": 0.00236
|
| 1364 |
+
},
|
| 1365 |
+
{
|
| 1366 |
+
"clip_ratio/high_max": 0.004156995622906834,
|
| 1367 |
+
"clip_ratio/high_mean": 0.004156995622906834,
|
| 1368 |
+
"clip_ratio/low_mean": 0.11334922257810831,
|
| 1369 |
+
"clip_ratio/low_min": 0.11334922257810831,
|
| 1370 |
+
"clip_ratio/region_mean": 0.11750621721148491,
|
| 1371 |
+
"epoch": 0.7977839335180056,
|
| 1372 |
+
"grad_norm": 1.5028809890146027,
|
| 1373 |
+
"kl": 3.652272095438093,
|
| 1374 |
+
"learning_rate": 0.00014016954246529696,
|
| 1375 |
+
"loss": 0.00900588370859623,
|
| 1376 |
+
"memory(GiB)": 80.03,
|
| 1377 |
+
"step": 72,
|
| 1378 |
+
"train_speed(iter/s)": 0.002378
|
| 1379 |
+
},
|
| 1380 |
+
{
|
| 1381 |
+
"clip_ratio/high_max": 0.0024562697508372366,
|
| 1382 |
+
"clip_ratio/high_mean": 0.0024562697508372366,
|
| 1383 |
+
"clip_ratio/low_mean": 0.00900937442202121,
|
| 1384 |
+
"clip_ratio/low_min": 0.00900937442202121,
|
| 1385 |
+
"clip_ratio/region_mean": 0.01146564440568909,
|
| 1386 |
+
"completions/clipped_ratio": 0.1015625,
|
| 1387 |
+
"completions/max_length": 101.0,
|
| 1388 |
+
"completions/mean_length": 20.9609375,
|
| 1389 |
+
"completions/min_length": 6.0,
|
| 1390 |
+
"epoch": 0.8088642659279779,
|
| 1391 |
+
"grad_norm": 1.9262088024122541,
|
| 1392 |
+
"kl": 2.068145776007441,
|
| 1393 |
+
"learning_rate": 0.00013848040976744457,
|
| 1394 |
+
"loss": 0.009554898366332054,
|
| 1395 |
+
"memory(GiB)": 80.03,
|
| 1396 |
+
"reward": 0.7734375,
|
| 1397 |
+
"reward_std": 0.6535134315490723,
|
| 1398 |
+
"rewards/Response_no_think_reward/mean": 0.7734375,
|
| 1399 |
+
"rewards/Response_no_think_reward/std": 1.224518895149231,
|
| 1400 |
+
"step": 73,
|
| 1401 |
+
"train_speed(iter/s)": 0.002347
|
| 1402 |
+
},
|
| 1403 |
+
{
|
| 1404 |
+
"clip_ratio/high_max": 0.05884167249314487,
|
| 1405 |
+
"clip_ratio/high_mean": 0.05884167249314487,
|
| 1406 |
+
"clip_ratio/low_mean": 0.014398490195162594,
|
| 1407 |
+
"clip_ratio/low_min": 0.014398490195162594,
|
| 1408 |
+
"clip_ratio/region_mean": 0.0732401623390615,
|
| 1409 |
+
"epoch": 0.8199445983379502,
|
| 1410 |
+
"grad_norm": 3.7794078536832525,
|
| 1411 |
+
"kl": 1.1005137297324836,
|
| 1412 |
+
"learning_rate": 0.00013677828928738934,
|
| 1413 |
+
"loss": 0.027932219207286835,
|
| 1414 |
+
"memory(GiB)": 80.03,
|
| 1415 |
+
"step": 74,
|
| 1416 |
+
"train_speed(iter/s)": 0.002365
|
| 1417 |
+
},
|
| 1418 |
+
{
|
| 1419 |
+
"clip_ratio/high_max": 0.0020199596765451133,
|
| 1420 |
+
"clip_ratio/high_mean": 0.0020199596765451133,
|
| 1421 |
+
"clip_ratio/low_mean": 0.009859619050985202,
|
| 1422 |
+
"clip_ratio/low_min": 0.009859619050985202,
|
| 1423 |
+
"clip_ratio/region_mean": 0.0118795787129784,
|
| 1424 |
+
"completions/clipped_ratio": 0.09375,
|
| 1425 |
+
"completions/max_length": 101.0,
|
| 1426 |
+
"completions/mean_length": 19.734375,
|
| 1427 |
+
"completions/min_length": 2.0,
|
| 1428 |
+
"epoch": 0.8310249307479224,
|
| 1429 |
+
"grad_norm": 2.6931881467895598,
|
| 1430 |
+
"kl": 1.9703011065721512,
|
| 1431 |
+
"learning_rate": 0.00013506375551927547,
|
| 1432 |
+
"loss": 0.010023066774010658,
|
| 1433 |
+
"memory(GiB)": 80.03,
|
| 1434 |
+
"reward": 0.4140625,
|
| 1435 |
+
"reward_std": 0.6804871559143066,
|
| 1436 |
+
"rewards/Response_no_think_reward/mean": 0.4140625,
|
| 1437 |
+
"rewards/Response_no_think_reward/std": 1.2005729675292969,
|
| 1438 |
+
"step": 75,
|
| 1439 |
+
"train_speed(iter/s)": 0.00233
|
| 1440 |
+
},
|
| 1441 |
+
{
|
| 1442 |
+
"clip_ratio/high_max": 0.004052987133036368,
|
| 1443 |
+
"clip_ratio/high_mean": 0.004052987133036368,
|
| 1444 |
+
"clip_ratio/low_mean": 0.14585177681874484,
|
| 1445 |
+
"clip_ratio/low_min": 0.14585177681874484,
|
| 1446 |
+
"clip_ratio/region_mean": 0.14990476449020207,
|
| 1447 |
+
"epoch": 0.8421052631578947,
|
| 1448 |
+
"grad_norm": 1.4459331642351634,
|
| 1449 |
+
"kl": 4.161040774546564,
|
| 1450 |
+
"learning_rate": 0.00013333738714693956,
|
| 1451 |
+
"loss": -0.005643587093800306,
|
| 1452 |
+
"memory(GiB)": 80.03,
|
| 1453 |
+
"step": 76,
|
| 1454 |
+
"train_speed(iter/s)": 0.002347
|
| 1455 |
+
},
|
| 1456 |
+
{
|
| 1457 |
+
"clip_ratio/high_max": 0.00046641789958812296,
|
| 1458 |
+
"clip_ratio/high_mean": 0.00046641789958812296,
|
| 1459 |
+
"clip_ratio/low_mean": 0.0038580247201025486,
|
| 1460 |
+
"clip_ratio/low_min": 0.0038580247201025486,
|
| 1461 |
+
"clip_ratio/region_mean": 0.0043244426196906716,
|
| 1462 |
+
"completions/clipped_ratio": 0.0859375,
|
| 1463 |
+
"completions/max_length": 99.0,
|
| 1464 |
+
"completions/mean_length": 14.8828125,
|
| 1465 |
+
"completions/min_length": 2.0,
|
| 1466 |
+
"epoch": 0.853185595567867,
|
| 1467 |
+
"grad_norm": 2.393092790871116,
|
| 1468 |
+
"kl": 2.344057558570057,
|
| 1469 |
+
"learning_rate": 0.00013159976684859527,
|
| 1470 |
+
"loss": 0.014792806468904018,
|
| 1471 |
+
"memory(GiB)": 80.03,
|
| 1472 |
+
"reward": 0.8046875,
|
| 1473 |
+
"reward_std": 0.515557050704956,
|
| 1474 |
+
"rewards/Response_no_think_reward/mean": 0.8046875,
|
| 1475 |
+
"rewards/Response_no_think_reward/std": 1.2862604856491089,
|
| 1476 |
+
"step": 77,
|
| 1477 |
+
"train_speed(iter/s)": 0.002292
|
| 1478 |
+
},
|
| 1479 |
+
{
|
| 1480 |
+
"clip_ratio/high_max": 0.005874009046237916,
|
| 1481 |
+
"clip_ratio/high_mean": 0.005874009046237916,
|
| 1482 |
+
"clip_ratio/low_mean": 0.0609364231931977,
|
| 1483 |
+
"clip_ratio/low_min": 0.0609364231931977,
|
| 1484 |
+
"clip_ratio/region_mean": 0.0668104327050969,
|
| 1485 |
+
"epoch": 0.8642659279778393,
|
| 1486 |
+
"grad_norm": 0.9235795825109938,
|
| 1487 |
+
"kl": 3.164612793829292,
|
| 1488 |
+
"learning_rate": 0.00012985148110016947,
|
| 1489 |
+
"loss": 0.001517204917035997,
|
| 1490 |
+
"memory(GiB)": 80.03,
|
| 1491 |
+
"step": 78,
|
| 1492 |
+
"train_speed(iter/s)": 0.002309
|
| 1493 |
+
},
|
| 1494 |
+
{
|
| 1495 |
+
"clip_ratio/high_max": 0.002277967141708359,
|
| 1496 |
+
"clip_ratio/high_mean": 0.002277967141708359,
|
| 1497 |
+
"clip_ratio/low_mean": 0.0037592062435578555,
|
| 1498 |
+
"clip_ratio/low_min": 0.0037592062435578555,
|
| 1499 |
+
"clip_ratio/region_mean": 0.006037173385266215,
|
| 1500 |
+
"completions/clipped_ratio": 0.0625,
|
| 1501 |
+
"completions/max_length": 87.0,
|
| 1502 |
+
"completions/mean_length": 15.703125,
|
| 1503 |
+
"completions/min_length": 2.0,
|
| 1504 |
+
"epoch": 0.8753462603878116,
|
| 1505 |
+
"grad_norm": 1.666555946903183,
|
| 1506 |
+
"kl": 2.2231151023879647,
|
| 1507 |
+
"learning_rate": 0.00012809311997735696,
|
| 1508 |
+
"loss": 0.012878447771072388,
|
| 1509 |
+
"memory(GiB)": 80.03,
|
| 1510 |
+
"reward": 0.546875,
|
| 1511 |
+
"reward_std": 0.7733994126319885,
|
| 1512 |
+
"rewards/Response_no_think_reward/mean": 0.546875,
|
| 1513 |
+
"rewards/Response_no_think_reward/std": 1.3032931089401245,
|
| 1514 |
+
"step": 79,
|
| 1515 |
+
"train_speed(iter/s)": 0.002271
|
| 1516 |
+
},
|
| 1517 |
+
{
|
| 1518 |
+
"clip_ratio/high_max": 0.03329824731918052,
|
| 1519 |
+
"clip_ratio/high_mean": 0.03329824731918052,
|
| 1520 |
+
"clip_ratio/low_mean": 0.017765316180884838,
|
| 1521 |
+
"clip_ratio/low_min": 0.017765316180884838,
|
| 1522 |
+
"clip_ratio/region_mean": 0.05106356361648068,
|
| 1523 |
+
"epoch": 0.8864265927977839,
|
| 1524 |
+
"grad_norm": 1.0568349375834465,
|
| 1525 |
+
"kl": 1.923786539278808,
|
| 1526 |
+
"learning_rate": 0.00012632527695645993,
|
| 1527 |
+
"loss": -0.0005231135291978717,
|
| 1528 |
+
"memory(GiB)": 80.03,
|
| 1529 |
+
"step": 80,
|
| 1530 |
+
"train_speed(iter/s)": 0.002287
|
| 1531 |
+
},
|
| 1532 |
+
{
|
| 1533 |
+
"clip_ratio/high_max": 0.00021551724057644606,
|
| 1534 |
+
"clip_ratio/high_mean": 0.00021551724057644606,
|
| 1535 |
+
"clip_ratio/low_mean": 0.006801646784879267,
|
| 1536 |
+
"clip_ratio/low_min": 0.006801646784879267,
|
| 1537 |
+
"clip_ratio/region_mean": 0.007017164025455713,
|
| 1538 |
+
"completions/clipped_ratio": 0.046875,
|
| 1539 |
+
"completions/max_length": 90.0,
|
| 1540 |
+
"completions/mean_length": 14.03125,
|
| 1541 |
+
"completions/min_length": 2.0,
|
| 1542 |
+
"epoch": 0.8975069252077562,
|
| 1543 |
+
"grad_norm": 9.477113742654934,
|
| 1544 |
+
"kl": 3.0111945159733295,
|
| 1545 |
+
"learning_rate": 0.00012454854871407994,
|
| 1546 |
+
"loss": 0.016959797590970993,
|
| 1547 |
+
"memory(GiB)": 80.03,
|
| 1548 |
+
"reward": 0.2578125,
|
| 1549 |
+
"reward_std": 0.5107755661010742,
|
| 1550 |
+
"rewards/Response_no_think_reward/mean": 0.2578125,
|
| 1551 |
+
"rewards/Response_no_think_reward/std": 1.2874077558517456,
|
| 1552 |
+
"step": 81,
|
| 1553 |
+
"train_speed(iter/s)": 0.002263
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"clip_ratio/high_max": 0.012387449765810743,
|
| 1557 |
+
"clip_ratio/high_mean": 0.012387449765810743,
|
| 1558 |
+
"clip_ratio/low_mean": 0.04806764563545585,
|
| 1559 |
+
"clip_ratio/low_min": 0.04806764563545585,
|
| 1560 |
+
"clip_ratio/region_mean": 0.06045509548857808,
|
| 1561 |
+
"epoch": 0.9085872576177285,
|
| 1562 |
+
"grad_norm": 1.1639747912015215,
|
| 1563 |
+
"kl": 3.7243148013949394,
|
| 1564 |
+
"learning_rate": 0.00012276353492572935,
|
| 1565 |
+
"loss": 0.0026693246327340603,
|
| 1566 |
+
"memory(GiB)": 80.03,
|
| 1567 |
+
"step": 82,
|
| 1568 |
+
"train_speed(iter/s)": 0.002279
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"clip_ratio/high_max": 0.0004032258002553135,
|
| 1572 |
+
"clip_ratio/high_mean": 0.0004032258002553135,
|
| 1573 |
+
"clip_ratio/low_mean": 0.0061736139468848705,
|
| 1574 |
+
"clip_ratio/low_min": 0.0061736139468848705,
|
| 1575 |
+
"clip_ratio/region_mean": 0.006576839747140184,
|
| 1576 |
+
"completions/clipped_ratio": 0.03125,
|
| 1577 |
+
"completions/max_length": 99.0,
|
| 1578 |
+
"completions/mean_length": 12.625,
|
| 1579 |
+
"completions/min_length": 2.0,
|
| 1580 |
+
"epoch": 0.9196675900277008,
|
| 1581 |
+
"grad_norm": 7.345333737975268,
|
| 1582 |
+
"kl": 1.6264500059187412,
|
| 1583 |
+
"learning_rate": 0.00012097083806343103,
|
| 1584 |
+
"loss": 0.02225027047097683,
|
| 1585 |
+
"memory(GiB)": 80.03,
|
| 1586 |
+
"reward": 0.6640625,
|
| 1587 |
+
"reward_std": 0.4172249436378479,
|
| 1588 |
+
"rewards/Response_no_think_reward/mean": 0.6640625,
|
| 1589 |
+
"rewards/Response_no_think_reward/std": 1.1588573455810547,
|
| 1590 |
+
"step": 83,
|
| 1591 |
+
"train_speed(iter/s)": 0.00227
|
| 1592 |
+
},
|
| 1593 |
+
{
|
| 1594 |
+
"clip_ratio/high_max": 0.0035271961241960526,
|
| 1595 |
+
"clip_ratio/high_mean": 0.0035271961241960526,
|
| 1596 |
+
"clip_ratio/low_mean": 0.02831025089835748,
|
| 1597 |
+
"clip_ratio/low_min": 0.02831025089835748,
|
| 1598 |
+
"clip_ratio/region_mean": 0.03183744702255353,
|
| 1599 |
+
"epoch": 0.9307479224376731,
|
| 1600 |
+
"grad_norm": 84.3494363093827,
|
| 1601 |
+
"kl": 10.616167868487537,
|
| 1602 |
+
"learning_rate": 0.00011917106319237386,
|
| 1603 |
+
"loss": 0.17172452807426453,
|
| 1604 |
+
"memory(GiB)": 80.03,
|
| 1605 |
+
"step": 84,
|
| 1606 |
+
"train_speed(iter/s)": 0.002286
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"clip_ratio/high_max": 0.0011701860348694026,
|
| 1610 |
+
"clip_ratio/high_mean": 0.0011701860348694026,
|
| 1611 |
+
"clip_ratio/low_mean": 0.014238029951229692,
|
| 1612 |
+
"clip_ratio/low_min": 0.014238029951229692,
|
| 1613 |
+
"clip_ratio/region_mean": 0.015408215986099094,
|
| 1614 |
+
"completions/clipped_ratio": 0.140625,
|
| 1615 |
+
"completions/max_length": 101.0,
|
| 1616 |
+
"completions/mean_length": 19.7578125,
|
| 1617 |
+
"completions/min_length": 4.0,
|
| 1618 |
+
"epoch": 0.9418282548476454,
|
| 1619 |
+
"grad_norm": 1.9907484000781221,
|
| 1620 |
+
"kl": 2.995624510163907,
|
| 1621 |
+
"learning_rate": 0.00011736481776669306,
|
| 1622 |
+
"loss": 0.01750401221215725,
|
| 1623 |
+
"memory(GiB)": 80.03,
|
| 1624 |
+
"reward": 0.4765625,
|
| 1625 |
+
"reward_std": 0.4854952394962311,
|
| 1626 |
+
"rewards/Response_no_think_reward/mean": 0.4765625,
|
| 1627 |
+
"rewards/Response_no_think_reward/std": 1.1080580949783325,
|
| 1628 |
+
"step": 85,
|
| 1629 |
+
"train_speed(iter/s)": 0.002274
|
| 1630 |
+
},
|
| 1631 |
+
{
|
| 1632 |
+
"clip_ratio/high_max": 0.008182557401596569,
|
| 1633 |
+
"clip_ratio/high_mean": 0.008182557401596569,
|
| 1634 |
+
"clip_ratio/low_mean": 0.11124554229900241,
|
| 1635 |
+
"clip_ratio/low_min": 0.11124554229900241,
|
| 1636 |
+
"clip_ratio/region_mean": 0.11942810016626026,
|
| 1637 |
+
"epoch": 0.9529085872576177,
|
| 1638 |
+
"grad_norm": 1.9564641583381899,
|
| 1639 |
+
"kl": 4.675610944104847,
|
| 1640 |
+
"learning_rate": 0.00011555271142444433,
|
| 1641 |
+
"loss": 0.009754904545843601,
|
| 1642 |
+
"memory(GiB)": 80.03,
|
| 1643 |
+
"step": 86,
|
| 1644 |
+
"train_speed(iter/s)": 0.002289
|
| 1645 |
+
},
|
| 1646 |
+
{
|
| 1647 |
+
"clip_ratio/high_max": 0.0004139889351790771,
|
| 1648 |
+
"clip_ratio/high_mean": 0.0004139889351790771,
|
| 1649 |
+
"clip_ratio/low_mean": 0.009557914454489946,
|
| 1650 |
+
"clip_ratio/low_min": 0.009557914454489946,
|
| 1651 |
+
"clip_ratio/region_mean": 0.0099719034624286,
|
| 1652 |
+
"completions/clipped_ratio": 0.0859375,
|
| 1653 |
+
"completions/max_length": 101.0,
|
| 1654 |
+
"completions/mean_length": 17.90625,
|
| 1655 |
+
"completions/min_length": 2.0,
|
| 1656 |
+
"epoch": 0.96398891966759,
|
| 1657 |
+
"grad_norm": 1.805689670049186,
|
| 1658 |
+
"kl": 3.4809365491382778,
|
| 1659 |
+
"learning_rate": 0.00011373535578184082,
|
| 1660 |
+
"loss": 0.0213579460978508,
|
| 1661 |
+
"memory(GiB)": 80.03,
|
| 1662 |
+
"reward": 0.5859375,
|
| 1663 |
+
"reward_std": 0.5313136577606201,
|
| 1664 |
+
"rewards/Response_no_think_reward/mean": 0.5859375,
|
| 1665 |
+
"rewards/Response_no_think_reward/std": 1.1939964294433594,
|
| 1666 |
+
"step": 87,
|
| 1667 |
+
"train_speed(iter/s)": 0.002277
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"clip_ratio/high_max": 0.009385127894347534,
|
| 1671 |
+
"clip_ratio/high_mean": 0.009385127894347534,
|
| 1672 |
+
"clip_ratio/low_mean": 0.08647578035015613,
|
| 1673 |
+
"clip_ratio/low_min": 0.08647578035015613,
|
| 1674 |
+
"clip_ratio/region_mean": 0.095860909903422,
|
| 1675 |
+
"epoch": 0.9750692520775623,
|
| 1676 |
+
"grad_norm": 2.411018383110019,
|
| 1677 |
+
"kl": 4.761912747140741,
|
| 1678 |
+
"learning_rate": 0.00011191336422682237,
|
| 1679 |
+
"loss": 0.01895134523510933,
|
| 1680 |
+
"memory(GiB)": 80.03,
|
| 1681 |
+
"step": 88,
|
| 1682 |
+
"train_speed(iter/s)": 0.002291
|
| 1683 |
+
},
|
| 1684 |
+
{
|
| 1685 |
+
"clip_ratio/high_max": 0.0031819915457163006,
|
| 1686 |
+
"clip_ratio/high_mean": 0.0031819915457163006,
|
| 1687 |
+
"clip_ratio/low_mean": 0.012216789647936821,
|
| 1688 |
+
"clip_ratio/low_min": 0.012216789647936821,
|
| 1689 |
+
"clip_ratio/region_mean": 0.015398780989926308,
|
| 1690 |
+
"completions/clipped_ratio": 0.0625,
|
| 1691 |
+
"completions/max_length": 64.0,
|
| 1692 |
+
"completions/mean_length": 11.28125,
|
| 1693 |
+
"completions/min_length": 2.0,
|
| 1694 |
+
"epoch": 0.9861495844875346,
|
| 1695 |
+
"grad_norm": 2.2626010793409574,
|
| 1696 |
+
"kl": 6.020935451146215,
|
| 1697 |
+
"learning_rate": 0.00011008735171202684,
|
| 1698 |
+
"loss": 0.031282562762498856,
|
| 1699 |
+
"memory(GiB)": 80.03,
|
| 1700 |
+
"reward": 0.2890625,
|
| 1701 |
+
"reward_std": 0.847247302532196,
|
| 1702 |
+
"rewards/Response_no_think_reward/mean": 0.2890625,
|
| 1703 |
+
"rewards/Response_no_think_reward/std": 1.1915208101272583,
|
| 1704 |
+
"step": 89,
|
| 1705 |
+
"train_speed(iter/s)": 0.002254
|
| 1706 |
+
},
|
| 1707 |
+
{
|
| 1708 |
+
"clip_ratio/high_max": 0.024421937006991357,
|
| 1709 |
+
"clip_ratio/high_mean": 0.024421937006991357,
|
| 1710 |
+
"clip_ratio/low_mean": 0.0521851975354366,
|
| 1711 |
+
"clip_ratio/low_min": 0.0521851975354366,
|
| 1712 |
+
"clip_ratio/region_mean": 0.07660713430959731,
|
| 1713 |
+
"epoch": 0.997229916897507,
|
| 1714 |
+
"grad_norm": 1.8302536272850396,
|
| 1715 |
+
"kl": 6.024846433196217,
|
| 1716 |
+
"learning_rate": 0.00010825793454723325,
|
| 1717 |
+
"loss": 0.017346249893307686,
|
| 1718 |
+
"memory(GiB)": 80.03,
|
| 1719 |
+
"step": 90,
|
| 1720 |
+
"train_speed(iter/s)": 0.002268
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"clip_ratio/high_max": 0.0011776478204410523,
|
| 1724 |
+
"clip_ratio/high_mean": 0.0011776478204410523,
|
| 1725 |
+
"clip_ratio/low_mean": 0.0035608798498287797,
|
| 1726 |
+
"clip_ratio/low_min": 0.0035608798498287797,
|
| 1727 |
+
"clip_ratio/region_mean": 0.004738527670269832,
|
| 1728 |
+
"completions/clipped_ratio": 0.125,
|
| 1729 |
+
"completions/max_length": 101.0,
|
| 1730 |
+
"completions/mean_length": 24.046875,
|
| 1731 |
+
"completions/min_length": 3.0,
|
| 1732 |
+
"epoch": 1.0110803324099722,
|
| 1733 |
+
"grad_norm": 2.3698869054826233,
|
| 1734 |
+
"kl": 5.333947827733937,
|
| 1735 |
+
"learning_rate": 0.00010642573019134703,
|
| 1736 |
+
"loss": 0.028505954891443253,
|
| 1737 |
+
"memory(GiB)": 80.03,
|
| 1738 |
+
"reward": 0.84375,
|
| 1739 |
+
"reward_std": 0.6655995845794678,
|
| 1740 |
+
"rewards/Response_no_think_reward/mean": 0.84375,
|
| 1741 |
+
"rewards/Response_no_think_reward/std": 1.2065274715423584,
|
| 1742 |
+
"step": 91,
|
| 1743 |
+
"train_speed(iter/s)": 0.002232
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"clip_ratio/high_max": 0.02221274602925405,
|
| 1747 |
+
"clip_ratio/high_mean": 0.02221274602925405,
|
| 1748 |
+
"clip_ratio/low_mean": 0.04682085904642008,
|
| 1749 |
+
"clip_ratio/low_min": 0.04682085904642008,
|
| 1750 |
+
"clip_ratio/region_mean": 0.0690336050465703,
|
| 1751 |
+
"epoch": 1.0221606648199446,
|
| 1752 |
+
"grad_norm": 1.377879165623118,
|
| 1753 |
+
"kl": 6.767704317186144,
|
| 1754 |
+
"learning_rate": 0.00010459135704399718,
|
| 1755 |
+
"loss": 0.01659482903778553,
|
| 1756 |
+
"memory(GiB)": 80.03,
|
| 1757 |
+
"step": 92,
|
| 1758 |
+
"train_speed(iter/s)": 0.002246
|
| 1759 |
+
},
|
| 1760 |
+
{
|
| 1761 |
+
"clip_ratio/high_max": 0.0007944914977997541,
|
| 1762 |
+
"clip_ratio/high_mean": 0.0007944914977997541,
|
| 1763 |
+
"clip_ratio/low_mean": 0.005214237666223198,
|
| 1764 |
+
"clip_ratio/low_min": 0.005214237666223198,
|
| 1765 |
+
"clip_ratio/region_mean": 0.006008729164022952,
|
| 1766 |
+
"completions/clipped_ratio": 0.046875,
|
| 1767 |
+
"completions/max_length": 79.0,
|
| 1768 |
+
"completions/mean_length": 11.1953125,
|
| 1769 |
+
"completions/min_length": 2.0,
|
| 1770 |
+
"epoch": 1.0332409972299168,
|
| 1771 |
+
"grad_norm": 1.128328936062302,
|
| 1772 |
+
"kl": 2.6099998716963455,
|
| 1773 |
+
"learning_rate": 0.00010275543423681621,
|
| 1774 |
+
"loss": 0.01443527452647686,
|
| 1775 |
+
"memory(GiB)": 80.03,
|
| 1776 |
+
"reward": 0.5703125,
|
| 1777 |
+
"reward_std": 0.34169840812683105,
|
| 1778 |
+
"rewards/Response_no_think_reward/mean": 0.5703125,
|
| 1779 |
+
"rewards/Response_no_think_reward/std": 1.26543128490448,
|
| 1780 |
+
"step": 93,
|
| 1781 |
+
"train_speed(iter/s)": 0.00223
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"clip_ratio/high_max": 0.024202606233302504,
|
| 1785 |
+
"clip_ratio/high_mean": 0.024202606233302504,
|
| 1786 |
+
"clip_ratio/low_mean": 0.017650849069468677,
|
| 1787 |
+
"clip_ratio/low_min": 0.017650849069468677,
|
| 1788 |
+
"clip_ratio/region_mean": 0.04185345536097884,
|
| 1789 |
+
"epoch": 1.0443213296398892,
|
| 1790 |
+
"grad_norm": 0.5979565041663649,
|
| 1791 |
+
"kl": 1.9247902451315895,
|
| 1792 |
+
"learning_rate": 0.00010091858142447265,
|
| 1793 |
+
"loss": 0.005481676664203405,
|
| 1794 |
+
"memory(GiB)": 80.03,
|
| 1795 |
+
"step": 94,
|
| 1796 |
+
"train_speed(iter/s)": 0.002243
|
| 1797 |
+
},
|
| 1798 |
+
{
|
| 1799 |
+
"clip_ratio/high_max": 0.002601411077193916,
|
| 1800 |
+
"clip_ratio/high_mean": 0.002601411077193916,
|
| 1801 |
+
"clip_ratio/low_mean": 0.005088170932140201,
|
| 1802 |
+
"clip_ratio/low_min": 0.005088170932140201,
|
| 1803 |
+
"clip_ratio/region_mean": 0.007689581951126456,
|
| 1804 |
+
"completions/clipped_ratio": 0.09375,
|
| 1805 |
+
"completions/max_length": 101.0,
|
| 1806 |
+
"completions/mean_length": 18.25,
|
| 1807 |
+
"completions/min_length": 4.0,
|
| 1808 |
+
"epoch": 1.0554016620498614,
|
| 1809 |
+
"grad_norm": 2.0648308983451917,
|
| 1810 |
+
"kl": 1.713204285595566,
|
| 1811 |
+
"learning_rate": 9.908141857552737e-05,
|
| 1812 |
+
"loss": 0.015094820410013199,
|
| 1813 |
+
"memory(GiB)": 80.03,
|
| 1814 |
+
"reward": 0.8046875,
|
| 1815 |
+
"reward_std": 0.47176384925842285,
|
| 1816 |
+
"rewards/Response_no_think_reward/mean": 0.8046875,
|
| 1817 |
+
"rewards/Response_no_think_reward/std": 1.1708977222442627,
|
| 1818 |
+
"step": 95,
|
| 1819 |
+
"train_speed(iter/s)": 0.002221
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"clip_ratio/high_max": 0.01881888063508086,
|
| 1823 |
+
"clip_ratio/high_mean": 0.01881888063508086,
|
| 1824 |
+
"clip_ratio/low_mean": 0.026617751631420106,
|
| 1825 |
+
"clip_ratio/low_min": 0.026617751631420106,
|
| 1826 |
+
"clip_ratio/region_mean": 0.0454366315389052,
|
| 1827 |
+
"epoch": 1.0664819944598338,
|
| 1828 |
+
"grad_norm": 1.5544793796965792,
|
| 1829 |
+
"kl": 1.6157679219031706,
|
| 1830 |
+
"learning_rate": 9.724456576318381e-05,
|
| 1831 |
+
"loss": 0.012739625759422779,
|
| 1832 |
+
"memory(GiB)": 80.03,
|
| 1833 |
+
"step": 96,
|
| 1834 |
+
"train_speed(iter/s)": 0.002234
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"clip_ratio/high_max": 0.0020559211261570454,
|
| 1838 |
+
"clip_ratio/high_mean": 0.0020559211261570454,
|
| 1839 |
+
"clip_ratio/low_mean": 0.0043267360888421535,
|
| 1840 |
+
"clip_ratio/low_min": 0.0043267360888421535,
|
| 1841 |
+
"clip_ratio/region_mean": 0.006382657098583877,
|
| 1842 |
+
"completions/clipped_ratio": 0.015625,
|
| 1843 |
+
"completions/max_length": 41.0,
|
| 1844 |
+
"completions/mean_length": 11.046875,
|
| 1845 |
+
"completions/min_length": 3.0,
|
| 1846 |
+
"epoch": 1.077562326869806,
|
| 1847 |
+
"grad_norm": 2.5531547462941115,
|
| 1848 |
+
"kl": 1.4817758100107312,
|
| 1849 |
+
"learning_rate": 9.540864295600283e-05,
|
| 1850 |
+
"loss": 0.016522858291864395,
|
| 1851 |
+
"memory(GiB)": 80.03,
|
| 1852 |
+
"reward": 0.9453125,
|
| 1853 |
+
"reward_std": 0.4848037362098694,
|
| 1854 |
+
"rewards/Response_no_think_reward/mean": 0.9453125,
|
| 1855 |
+
"rewards/Response_no_think_reward/std": 1.1454023122787476,
|
| 1856 |
+
"step": 97,
|
| 1857 |
+
"train_speed(iter/s)": 0.00222
|
| 1858 |
+
},
|
| 1859 |
+
{
|
| 1860 |
+
"clip_ratio/high_max": 0.01358306163456291,
|
| 1861 |
+
"clip_ratio/high_mean": 0.01358306163456291,
|
| 1862 |
+
"clip_ratio/low_mean": 0.019928007503040135,
|
| 1863 |
+
"clip_ratio/low_min": 0.019928007503040135,
|
| 1864 |
+
"clip_ratio/region_mean": 0.03351106960326433,
|
| 1865 |
+
"epoch": 1.0886426592797784,
|
| 1866 |
+
"grad_norm": 1.0799722223140724,
|
| 1867 |
+
"kl": 1.3886900492943823,
|
| 1868 |
+
"learning_rate": 9.357426980865301e-05,
|
| 1869 |
+
"loss": -0.0007159767556004226,
|
| 1870 |
+
"memory(GiB)": 80.03,
|
| 1871 |
+
"step": 98,
|
| 1872 |
+
"train_speed(iter/s)": 0.002234
|
| 1873 |
+
},
|
| 1874 |
+
{
|
| 1875 |
+
"clip_ratio/high_max": 0.001544758939417079,
|
| 1876 |
+
"clip_ratio/high_mean": 0.001544758939417079,
|
| 1877 |
+
"clip_ratio/low_mean": 0.0037943847200949676,
|
| 1878 |
+
"clip_ratio/low_min": 0.0037943847200949676,
|
| 1879 |
+
"clip_ratio/region_mean": 0.005339143652236089,
|
| 1880 |
+
"completions/clipped_ratio": 0.0546875,
|
| 1881 |
+
"completions/max_length": 101.0,
|
| 1882 |
+
"completions/mean_length": 18.546875,
|
| 1883 |
+
"completions/min_length": 3.0,
|
| 1884 |
+
"epoch": 1.0997229916897506,
|
| 1885 |
+
"grad_norm": 0.943375664401446,
|
| 1886 |
+
"kl": 1.7392279328778386,
|
| 1887 |
+
"learning_rate": 9.174206545276677e-05,
|
| 1888 |
+
"loss": 0.014223725534975529,
|
| 1889 |
+
"memory(GiB)": 80.03,
|
| 1890 |
+
"reward": 1.1015625,
|
| 1891 |
+
"reward_std": 0.39179152250289917,
|
| 1892 |
+
"rewards/Response_no_think_reward/mean": 1.1015625,
|
| 1893 |
+
"rewards/Response_no_think_reward/std": 1.2221051454544067,
|
| 1894 |
+
"step": 99,
|
| 1895 |
+
"train_speed(iter/s)": 0.002219
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"clip_ratio/high_max": 0.01209607784403488,
|
| 1899 |
+
"clip_ratio/high_mean": 0.01209607784403488,
|
| 1900 |
+
"clip_ratio/low_mean": 0.011367922488716431,
|
| 1901 |
+
"clip_ratio/low_min": 0.011367922488716431,
|
| 1902 |
+
"clip_ratio/region_mean": 0.023464000318199396,
|
| 1903 |
+
"epoch": 1.110803324099723,
|
| 1904 |
+
"grad_norm": 0.8253425068817393,
|
| 1905 |
+
"kl": 1.7898913251701742,
|
| 1906 |
+
"learning_rate": 8.991264828797319e-05,
|
| 1907 |
+
"loss": 0.006411677226424217,
|
| 1908 |
+
"memory(GiB)": 80.03,
|
| 1909 |
+
"step": 100,
|
| 1910 |
+
"train_speed(iter/s)": 0.002232
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"clip_ratio/high_max": 0.0038593837525695562,
|
| 1914 |
+
"clip_ratio/high_mean": 0.0038593837525695562,
|
| 1915 |
+
"clip_ratio/low_mean": 0.007897521747509018,
|
| 1916 |
+
"clip_ratio/low_min": 0.007897521747509018,
|
| 1917 |
+
"clip_ratio/region_mean": 0.011756905500078574,
|
| 1918 |
+
"completions/clipped_ratio": 0.140625,
|
| 1919 |
+
"completions/max_length": 51.0,
|
| 1920 |
+
"completions/mean_length": 18.0390625,
|
| 1921 |
+
"completions/min_length": 2.0,
|
| 1922 |
+
"epoch": 1.1218836565096952,
|
| 1923 |
+
"grad_norm": 2.1091313211648033,
|
| 1924 |
+
"kl": 4.433779507409781,
|
| 1925 |
+
"learning_rate": 8.808663577317764e-05,
|
| 1926 |
+
"loss": 0.026584401726722717,
|
| 1927 |
+
"memory(GiB)": 77.36,
|
| 1928 |
+
"reward": 0.234375,
|
| 1929 |
+
"reward_std": 0.353938490152359,
|
| 1930 |
+
"rewards/Response_no_think_reward_1/mean": 0.234375,
|
| 1931 |
+
"rewards/Response_no_think_reward_1/std": 0.8275223970413208,
|
| 1932 |
+
"step": 101,
|
| 1933 |
+
"train_speed(iter/s)": 0.112917
|
| 1934 |
+
},
|
| 1935 |
+
{
|
| 1936 |
+
"clip_ratio/high_max": 0.0257808348396793,
|
| 1937 |
+
"clip_ratio/high_mean": 0.0257808348396793,
|
| 1938 |
+
"clip_ratio/low_mean": 0.013182859780499712,
|
| 1939 |
+
"clip_ratio/low_min": 0.013182859780499712,
|
| 1940 |
+
"clip_ratio/region_mean": 0.038963694794801995,
|
| 1941 |
+
"epoch": 1.1329639889196677,
|
| 1942 |
+
"grad_norm": 0.8069647208167382,
|
| 1943 |
+
"kl": 2.7774715912528336,
|
| 1944 |
+
"learning_rate": 8.626464421815919e-05,
|
| 1945 |
+
"loss": 0.008937789127230644,
|
| 1946 |
+
"memory(GiB)": 77.36,
|
| 1947 |
+
"step": 102,
|
| 1948 |
+
"train_speed(iter/s)": 0.09385
|
| 1949 |
+
},
|
| 1950 |
+
{
|
| 1951 |
+
"clip_ratio/high_max": 0.00028669723542407155,
|
| 1952 |
+
"clip_ratio/high_mean": 0.00028669723542407155,
|
| 1953 |
+
"clip_ratio/low_mean": 0.0008802816737443209,
|
| 1954 |
+
"clip_ratio/low_min": 0.0008802816737443209,
|
| 1955 |
+
"clip_ratio/region_mean": 0.0011669789091683924,
|
| 1956 |
+
"completions/clipped_ratio": 0.109375,
|
| 1957 |
+
"completions/max_length": 51.0,
|
| 1958 |
+
"completions/mean_length": 18.140625,
|
| 1959 |
+
"completions/min_length": 3.0,
|
| 1960 |
+
"epoch": 1.1440443213296398,
|
| 1961 |
+
"grad_norm": 3.4014714828505634,
|
| 1962 |
+
"kl": 1.8993340344168246,
|
| 1963 |
+
"learning_rate": 8.444728857555572e-05,
|
| 1964 |
+
"loss": 0.029605647549033165,
|
| 1965 |
+
"memory(GiB)": 77.47,
|
| 1966 |
+
"reward": 0.390625,
|
| 1967 |
+
"reward_std": 0.1173202246427536,
|
| 1968 |
+
"rewards/Response_no_think_reward_1/mean": 0.390625,
|
| 1969 |
+
"rewards/Response_no_think_reward_1/std": 0.8440096974372864,
|
| 1970 |
+
"step": 103,
|
| 1971 |
+
"train_speed(iter/s)": 0.053798
|
| 1972 |
+
},
|
| 1973 |
+
{
|
| 1974 |
+
"clip_ratio/high_max": 0.0013847328373230994,
|
| 1975 |
+
"clip_ratio/high_mean": 0.0013847328373230994,
|
| 1976 |
+
"clip_ratio/low_mean": 0.004156515002250671,
|
| 1977 |
+
"clip_ratio/low_min": 0.004156515002250671,
|
| 1978 |
+
"clip_ratio/region_mean": 0.00554124778136611,
|
| 1979 |
+
"epoch": 1.1551246537396123,
|
| 1980 |
+
"grad_norm": 0.6574286317590833,
|
| 1981 |
+
"kl": 0.8905834904871881,
|
| 1982 |
+
"learning_rate": 8.263518223330697e-05,
|
| 1983 |
+
"loss": 0.007299074437469244,
|
| 1984 |
+
"memory(GiB)": 77.47,
|
| 1985 |
+
"step": 104,
|
| 1986 |
+
"train_speed(iter/s)": 0.049358
|
| 1987 |
+
},
|
| 1988 |
+
{
|
| 1989 |
+
"clip_ratio/high_max": 0.00036549707874655724,
|
| 1990 |
+
"clip_ratio/high_mean": 0.00036549707874655724,
|
| 1991 |
+
"clip_ratio/low_mean": 0.003542276710504666,
|
| 1992 |
+
"clip_ratio/low_min": 0.003542276710504666,
|
| 1993 |
+
"clip_ratio/region_mean": 0.003907773789251223,
|
| 1994 |
+
"completions/clipped_ratio": 0.0546875,
|
| 1995 |
+
"completions/max_length": 51.0,
|
| 1996 |
+
"completions/mean_length": 12.1015625,
|
| 1997 |
+
"completions/min_length": 2.0,
|
| 1998 |
+
"epoch": 1.1662049861495845,
|
| 1999 |
+
"grad_norm": 1.418255621232663,
|
| 2000 |
+
"kl": 1.0295969531871378,
|
| 2001 |
+
"learning_rate": 8.082893680762619e-05,
|
| 2002 |
+
"loss": 0.008070322684943676,
|
| 2003 |
+
"memory(GiB)": 77.47,
|
| 2004 |
+
"reward": 0.5625,
|
| 2005 |
+
"reward_std": 0.283821702003479,
|
| 2006 |
+
"rewards/Response_no_think_reward_1/mean": 0.5625,
|
| 2007 |
+
"rewards/Response_no_think_reward_1/std": 0.6728714108467102,
|
| 2008 |
+
"step": 105,
|
| 2009 |
+
"train_speed(iter/s)": 0.035507
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"clip_ratio/high_max": 0.015719514689408243,
|
| 2013 |
+
"clip_ratio/high_mean": 0.015719514689408243,
|
| 2014 |
+
"clip_ratio/low_mean": 0.010580109432339668,
|
| 2015 |
+
"clip_ratio/low_min": 0.010580109432339668,
|
| 2016 |
+
"clip_ratio/region_mean": 0.02629962412174791,
|
| 2017 |
+
"epoch": 1.1772853185595569,
|
| 2018 |
+
"grad_norm": 0.9124403050854863,
|
| 2019 |
+
"kl": 0.933376073371619,
|
| 2020 |
+
"learning_rate": 7.902916193656898e-05,
|
| 2021 |
+
"loss": -0.002763347467407584,
|
| 2022 |
+
"memory(GiB)": 79.43,
|
| 2023 |
+
"step": 106,
|
| 2024 |
+
"train_speed(iter/s)": 0.033633
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"clip_ratio/high_max": 0.0023960003745742142,
|
| 2028 |
+
"clip_ratio/high_mean": 0.0023960003745742142,
|
| 2029 |
+
"clip_ratio/low_mean": 0.002585217938758433,
|
| 2030 |
+
"clip_ratio/low_min": 0.002585217938758433,
|
| 2031 |
+
"clip_ratio/region_mean": 0.004981218371540308,
|
| 2032 |
+
"completions/clipped_ratio": 0.0625,
|
| 2033 |
+
"completions/max_length": 51.0,
|
| 2034 |
+
"completions/mean_length": 13.6328125,
|
| 2035 |
+
"completions/min_length": 2.0,
|
| 2036 |
+
"epoch": 1.188365650969529,
|
| 2037 |
+
"grad_norm": 2.0082146476980807,
|
| 2038 |
+
"kl": 1.3074679019264295,
|
| 2039 |
+
"learning_rate": 7.72364650742707e-05,
|
| 2040 |
+
"loss": 0.011914699338376522,
|
| 2041 |
+
"memory(GiB)": 79.43,
|
| 2042 |
+
"reward": 0.328125,
|
| 2043 |
+
"reward_std": 0.18394747376441956,
|
| 2044 |
+
"rewards/Response_no_think_reward_1/mean": 0.328125,
|
| 2045 |
+
"rewards/Response_no_think_reward_1/std": 0.7110973596572876,
|
| 2046 |
+
"step": 107,
|
| 2047 |
+
"train_speed(iter/s)": 0.027655
|
| 2048 |
+
},
|
| 2049 |
+
{
|
| 2050 |
+
"clip_ratio/high_max": 0.00717889575753361,
|
| 2051 |
+
"clip_ratio/high_mean": 0.00717889575753361,
|
| 2052 |
+
"clip_ratio/low_mean": 0.008150914101861417,
|
| 2053 |
+
"clip_ratio/low_min": 0.008150914101861417,
|
| 2054 |
+
"clip_ratio/region_mean": 0.015329810208640993,
|
| 2055 |
+
"epoch": 1.1994459833795015,
|
| 2056 |
+
"grad_norm": 0.5916941576021421,
|
| 2057 |
+
"kl": 1.0833495813399168,
|
| 2058 |
+
"learning_rate": 7.54514512859201e-05,
|
| 2059 |
+
"loss": 0.0037035662680864334,
|
| 2060 |
+
"memory(GiB)": 79.43,
|
| 2061 |
+
"step": 108,
|
| 2062 |
+
"train_speed(iter/s)": 0.026591
|
| 2063 |
+
},
|
| 2064 |
+
{
|
| 2065 |
+
"clip_ratio/high_max": 0.004251995822414756,
|
| 2066 |
+
"clip_ratio/high_mean": 0.004251995822414756,
|
| 2067 |
+
"clip_ratio/low_mean": 0.008190131688024849,
|
| 2068 |
+
"clip_ratio/low_min": 0.008190131688024849,
|
| 2069 |
+
"clip_ratio/region_mean": 0.012442127510439605,
|
| 2070 |
+
"completions/clipped_ratio": 0.09375,
|
| 2071 |
+
"completions/max_length": 51.0,
|
| 2072 |
+
"completions/mean_length": 12.234375,
|
| 2073 |
+
"completions/min_length": 4.0,
|
| 2074 |
+
"epoch": 1.2105263157894737,
|
| 2075 |
+
"grad_norm": 2.6396182071410137,
|
| 2076 |
+
"kl": 2.141993957106024,
|
| 2077 |
+
"learning_rate": 7.36747230435401e-05,
|
| 2078 |
+
"loss": 0.011151588521897793,
|
| 2079 |
+
"memory(GiB)": 79.43,
|
| 2080 |
+
"reward": -0.2265625,
|
| 2081 |
+
"reward_std": 0.3407740592956543,
|
| 2082 |
+
"rewards/Response_no_think_reward_1/mean": -0.2265625,
|
| 2083 |
+
"rewards/Response_no_think_reward_1/std": 0.6425201892852783,
|
| 2084 |
+
"step": 109,
|
| 2085 |
+
"train_speed(iter/s)": 0.022574
|
| 2086 |
+
},
|
| 2087 |
+
{
|
| 2088 |
+
"clip_ratio/high_max": 0.004035328107420355,
|
| 2089 |
+
"clip_ratio/high_mean": 0.004035328107420355,
|
| 2090 |
+
"clip_ratio/low_mean": 0.08847818686626852,
|
| 2091 |
+
"clip_ratio/low_min": 0.08847818686626852,
|
| 2092 |
+
"clip_ratio/region_mean": 0.0925135153811425,
|
| 2093 |
+
"epoch": 1.221606648199446,
|
| 2094 |
+
"grad_norm": 1.4171762109533736,
|
| 2095 |
+
"kl": 3.1860878374427557,
|
| 2096 |
+
"learning_rate": 7.190688002264308e-05,
|
| 2097 |
+
"loss": 0.0037668771110475063,
|
| 2098 |
+
"memory(GiB)": 79.43,
|
| 2099 |
+
"step": 110,
|
| 2100 |
+
"train_speed(iter/s)": 0.021907
|
| 2101 |
+
},
|
| 2102 |
+
{
|
| 2103 |
+
"clip_ratio/high_max": 0.0,
|
| 2104 |
+
"clip_ratio/high_mean": 0.0,
|
| 2105 |
+
"clip_ratio/low_mean": 0.002568919211626053,
|
| 2106 |
+
"clip_ratio/low_min": 0.002568919211626053,
|
| 2107 |
+
"clip_ratio/region_mean": 0.002568919211626053,
|
| 2108 |
+
"completions/clipped_ratio": 0.046875,
|
| 2109 |
+
"completions/max_length": 48.0,
|
| 2110 |
+
"completions/mean_length": 11.7734375,
|
| 2111 |
+
"completions/min_length": 5.0,
|
| 2112 |
+
"epoch": 1.2326869806094183,
|
| 2113 |
+
"grad_norm": 1.02177770841713,
|
| 2114 |
+
"kl": 1.6274185269139707,
|
| 2115 |
+
"learning_rate": 7.014851889983057e-05,
|
| 2116 |
+
"loss": 0.010478168725967407,
|
| 2117 |
+
"memory(GiB)": 79.43,
|
| 2118 |
+
"reward": 0.453125,
|
| 2119 |
+
"reward_std": 0.3291260004043579,
|
| 2120 |
+
"rewards/Response_no_think_reward_1/mean": 0.453125,
|
| 2121 |
+
"rewards/Response_no_think_reward_1/std": 0.685730516910553,
|
| 2122 |
+
"step": 111,
|
| 2123 |
+
"train_speed(iter/s)": 0.019127
|
| 2124 |
+
},
|
| 2125 |
+
{
|
| 2126 |
+
"clip_ratio/high_max": 0.013096909533487633,
|
| 2127 |
+
"clip_ratio/high_mean": 0.013096909533487633,
|
| 2128 |
+
"clip_ratio/low_mean": 0.016107605304569006,
|
| 2129 |
+
"clip_ratio/low_min": 0.016107605304569006,
|
| 2130 |
+
"clip_ratio/region_mean": 0.029204514692537487,
|
| 2131 |
+
"epoch": 1.2437673130193905,
|
| 2132 |
+
"grad_norm": 0.79999454502468,
|
| 2133 |
+
"kl": 1.5544351362623274,
|
| 2134 |
+
"learning_rate": 6.840023315140475e-05,
|
| 2135 |
+
"loss": 0.0022247314918786287,
|
| 2136 |
+
"memory(GiB)": 79.43,
|
| 2137 |
+
"step": 112,
|
| 2138 |
+
"train_speed(iter/s)": 0.018681
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"clip_ratio/high_max": 0.0008340688509633765,
|
| 2142 |
+
"clip_ratio/high_mean": 0.0008340688509633765,
|
| 2143 |
+
"clip_ratio/low_mean": 0.007082634954713285,
|
| 2144 |
+
"clip_ratio/low_min": 0.007082634954713285,
|
| 2145 |
+
"clip_ratio/region_mean": 0.007916703820228577,
|
| 2146 |
+
"completions/clipped_ratio": 0.1171875,
|
| 2147 |
+
"completions/max_length": 51.0,
|
| 2148 |
+
"completions/mean_length": 13.53125,
|
| 2149 |
+
"completions/min_length": 3.0,
|
| 2150 |
+
"epoch": 1.254847645429363,
|
| 2151 |
+
"grad_norm": 1.6246389101003882,
|
| 2152 |
+
"kl": 1.6449745513964444,
|
| 2153 |
+
"learning_rate": 6.666261285306047e-05,
|
| 2154 |
+
"loss": 0.014917208813130856,
|
| 2155 |
+
"memory(GiB)": 79.43,
|
| 2156 |
+
"reward": 0.1953125,
|
| 2157 |
+
"reward_std": 0.2754020392894745,
|
| 2158 |
+
"rewards/Response_no_think_reward_1/mean": 0.1953125,
|
| 2159 |
+
"rewards/Response_no_think_reward_1/std": 0.869958758354187,
|
| 2160 |
+
"step": 113,
|
| 2161 |
+
"train_speed(iter/s)": 0.016592
|
| 2162 |
+
},
|
| 2163 |
+
{
|
| 2164 |
+
"clip_ratio/high_max": 0.006006928611896001,
|
| 2165 |
+
"clip_ratio/high_mean": 0.006006928611896001,
|
| 2166 |
+
"clip_ratio/low_mean": 0.039706501411274076,
|
| 2167 |
+
"clip_ratio/low_min": 0.039706501411274076,
|
| 2168 |
+
"clip_ratio/region_mean": 0.045713430270552635,
|
| 2169 |
+
"epoch": 1.2659279778393353,
|
| 2170 |
+
"grad_norm": 0.940807595373605,
|
| 2171 |
+
"kl": 1.8106578167062253,
|
| 2172 |
+
"learning_rate": 6.493624448072457e-05,
|
| 2173 |
+
"loss": 0.006363555323332548,
|
| 2174 |
+
"memory(GiB)": 79.43,
|
| 2175 |
+
"step": 114,
|
| 2176 |
+
"train_speed(iter/s)": 0.016278
|
| 2177 |
+
},
|
| 2178 |
+
{
|
| 2179 |
+
"clip_ratio/high_max": 0.0052188277477398515,
|
| 2180 |
+
"clip_ratio/high_mean": 0.0052188277477398515,
|
| 2181 |
+
"clip_ratio/low_mean": 0.004236701555782929,
|
| 2182 |
+
"clip_ratio/low_min": 0.004236701555782929,
|
| 2183 |
+
"clip_ratio/region_mean": 0.009455529390834272,
|
| 2184 |
+
"completions/clipped_ratio": 0.1796875,
|
| 2185 |
+
"completions/max_length": 51.0,
|
| 2186 |
+
"completions/mean_length": 14.3671875,
|
| 2187 |
+
"completions/min_length": 2.0,
|
| 2188 |
+
"epoch": 1.2770083102493075,
|
| 2189 |
+
"grad_norm": 2.8040352101197437,
|
| 2190 |
+
"kl": 5.506896490347572,
|
| 2191 |
+
"learning_rate": 6.322171071261071e-05,
|
| 2192 |
+
"loss": 0.03229736536741257,
|
| 2193 |
+
"memory(GiB)": 79.43,
|
| 2194 |
+
"reward": 0.296875,
|
| 2195 |
+
"reward_std": 0.43980443477630615,
|
| 2196 |
+
"rewards/Response_no_think_reward_1/mean": 0.296875,
|
| 2197 |
+
"rewards/Response_no_think_reward_1/std": 0.8905397057533264,
|
| 2198 |
+
"step": 115,
|
| 2199 |
+
"train_speed(iter/s)": 0.014636
|
| 2200 |
+
},
|
| 2201 |
+
{
|
| 2202 |
+
"clip_ratio/high_max": 0.01615347486222163,
|
| 2203 |
+
"clip_ratio/high_mean": 0.01615347486222163,
|
| 2204 |
+
"clip_ratio/low_mean": 0.015428576618432999,
|
| 2205 |
+
"clip_ratio/low_min": 0.015428576618432999,
|
| 2206 |
+
"clip_ratio/region_mean": 0.031582050723955035,
|
| 2207 |
+
"epoch": 1.2880886426592797,
|
| 2208 |
+
"grad_norm": 1.8650893219420759,
|
| 2209 |
+
"kl": 4.258469146443531,
|
| 2210 |
+
"learning_rate": 6.151959023255545e-05,
|
| 2211 |
+
"loss": 0.020270783454179764,
|
| 2212 |
+
"memory(GiB)": 79.43,
|
| 2213 |
+
"step": 116,
|
| 2214 |
+
"train_speed(iter/s)": 0.014407
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"clip_ratio/high_max": 0.005145640461705625,
|
| 2218 |
+
"clip_ratio/high_mean": 0.005145640461705625,
|
| 2219 |
+
"clip_ratio/low_mean": 0.008849590638419613,
|
| 2220 |
+
"clip_ratio/low_min": 0.008849590638419613,
|
| 2221 |
+
"clip_ratio/region_mean": 0.013995230983709916,
|
| 2222 |
+
"completions/clipped_ratio": 0.1640625,
|
| 2223 |
+
"completions/max_length": 51.0,
|
| 2224 |
+
"completions/mean_length": 16.3125,
|
| 2225 |
+
"completions/min_length": 2.0,
|
| 2226 |
+
"epoch": 1.299168975069252,
|
| 2227 |
+
"grad_norm": 2.4581301952944186,
|
| 2228 |
+
"kl": 3.745300827547908,
|
| 2229 |
+
"learning_rate": 5.983045753470308e-05,
|
| 2230 |
+
"loss": 0.024165078997612,
|
| 2231 |
+
"memory(GiB)": 79.43,
|
| 2232 |
+
"reward": 0.453125,
|
| 2233 |
+
"reward_std": 0.5826787948608398,
|
| 2234 |
+
"rewards/Response_no_think_reward_1/mean": 0.453125,
|
| 2235 |
+
"rewards/Response_no_think_reward_1/std": 0.8405039310455322,
|
| 2236 |
+
"step": 117,
|
| 2237 |
+
"train_speed(iter/s)": 0.013208
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"clip_ratio/high_max": 0.04523819196037948,
|
| 2241 |
+
"clip_ratio/high_mean": 0.04523819196037948,
|
| 2242 |
+
"clip_ratio/low_mean": 0.01510127488290891,
|
| 2243 |
+
"clip_ratio/low_min": 0.01510127488290891,
|
| 2244 |
+
"clip_ratio/region_mean": 0.060339466377627105,
|
| 2245 |
+
"epoch": 1.3102493074792243,
|
| 2246 |
+
"grad_norm": 2.131099568305027,
|
| 2247 |
+
"kl": 2.847630614414811,
|
| 2248 |
+
"learning_rate": 5.8154882729603876e-05,
|
| 2249 |
+
"loss": 0.01335166022181511,
|
| 2250 |
+
"memory(GiB)": 79.43,
|
| 2251 |
+
"step": 118,
|
| 2252 |
+
"train_speed(iter/s)": 0.013037
|
| 2253 |
+
},
|
| 2254 |
+
{
|
| 2255 |
+
"clip_ratio/high_max": 0.002125054510543123,
|
| 2256 |
+
"clip_ratio/high_mean": 0.002125054510543123,
|
| 2257 |
+
"clip_ratio/low_mean": 0.0127813016588334,
|
| 2258 |
+
"clip_ratio/low_min": 0.0127813016588334,
|
| 2259 |
+
"clip_ratio/region_mean": 0.0149063560529612,
|
| 2260 |
+
"completions/clipped_ratio": 0.203125,
|
| 2261 |
+
"completions/max_length": 51.0,
|
| 2262 |
+
"completions/mean_length": 19.59375,
|
| 2263 |
+
"completions/min_length": 3.0,
|
| 2264 |
+
"epoch": 1.3213296398891967,
|
| 2265 |
+
"grad_norm": 2.690385409772627,
|
| 2266 |
+
"kl": 2.0104650848079473,
|
| 2267 |
+
"learning_rate": 5.64934313517927e-05,
|
| 2268 |
+
"loss": 0.01620793715119362,
|
| 2269 |
+
"memory(GiB)": 79.43,
|
| 2270 |
+
"reward": 0.2578125,
|
| 2271 |
+
"reward_std": 0.41504764556884766,
|
| 2272 |
+
"rewards/Response_no_think_reward_1/mean": 0.2578125,
|
| 2273 |
+
"rewards/Response_no_think_reward_1/std": 0.8625734448432922,
|
| 2274 |
+
"step": 119,
|
| 2275 |
+
"train_speed(iter/s)": 0.011541
|
| 2276 |
+
},
|
| 2277 |
+
{
|
| 2278 |
+
"clip_ratio/high_max": 0.0026382115320302546,
|
| 2279 |
+
"clip_ratio/high_mean": 0.0026382115320302546,
|
| 2280 |
+
"clip_ratio/low_mean": 0.07074824426672421,
|
| 2281 |
+
"clip_ratio/low_min": 0.07074824426672421,
|
| 2282 |
+
"clip_ratio/region_mean": 0.07338645646814257,
|
| 2283 |
+
"epoch": 1.332409972299169,
|
| 2284 |
+
"grad_norm": 1.3611711034111147,
|
| 2285 |
+
"kl": 2.366683575557545,
|
| 2286 |
+
"learning_rate": 5.484666416891109e-05,
|
| 2287 |
+
"loss": 0.00427972199395299,
|
| 2288 |
+
"memory(GiB)": 79.43,
|
| 2289 |
+
"step": 120,
|
| 2290 |
+
"train_speed(iter/s)": 0.011424
|
| 2291 |
+
},
|
| 2292 |
+
{
|
| 2293 |
+
"clip_ratio/high_max": 0.00034340660204179585,
|
| 2294 |
+
"clip_ratio/high_mean": 0.00034340660204179585,
|
| 2295 |
+
"clip_ratio/low_mean": 0.0057576168910600245,
|
| 2296 |
+
"clip_ratio/low_min": 0.0057576168910600245,
|
| 2297 |
+
"clip_ratio/region_mean": 0.00610102349310182,
|
| 2298 |
+
"completions/clipped_ratio": 0.109375,
|
| 2299 |
+
"completions/max_length": 51.0,
|
| 2300 |
+
"completions/mean_length": 14.1640625,
|
| 2301 |
+
"completions/min_length": 3.0,
|
| 2302 |
+
"epoch": 1.3434903047091413,
|
| 2303 |
+
"grad_norm": 1.7730415769710828,
|
| 2304 |
+
"kl": 2.253111455589533,
|
| 2305 |
+
"learning_rate": 5.321513699243924e-05,
|
| 2306 |
+
"loss": 0.015828199684619904,
|
| 2307 |
+
"memory(GiB)": 79.43,
|
| 2308 |
+
"reward": 0.2578125,
|
| 2309 |
+
"reward_std": 0.2504267692565918,
|
| 2310 |
+
"rewards/Response_no_think_reward_1/mean": 0.2578125,
|
| 2311 |
+
"rewards/Response_no_think_reward_1/std": 0.8533961176872253,
|
| 2312 |
+
"step": 121,
|
| 2313 |
+
"train_speed(iter/s)": 0.010746
|
| 2314 |
+
},
|
| 2315 |
+
{
|
| 2316 |
+
"clip_ratio/high_max": 0.002522797236451879,
|
| 2317 |
+
"clip_ratio/high_mean": 0.002522797236451879,
|
| 2318 |
+
"clip_ratio/low_mean": 0.03892370511312038,
|
| 2319 |
+
"clip_ratio/low_min": 0.03892370511312038,
|
| 2320 |
+
"clip_ratio/region_mean": 0.041446502320468426,
|
| 2321 |
+
"epoch": 1.3545706371191135,
|
| 2322 |
+
"grad_norm": 0.9270755733265906,
|
| 2323 |
+
"kl": 2.852388353087008,
|
| 2324 |
+
"learning_rate": 5.159940049010015e-05,
|
| 2325 |
+
"loss": 0.00514911487698555,
|
| 2326 |
+
"memory(GiB)": 79.43,
|
| 2327 |
+
"step": 122,
|
| 2328 |
+
"train_speed(iter/s)": 0.010652
|
| 2329 |
+
},
|
| 2330 |
+
{
|
| 2331 |
+
"clip_ratio/high_max": 0.0008753835718380287,
|
| 2332 |
+
"clip_ratio/high_mean": 0.0008753835718380287,
|
| 2333 |
+
"clip_ratio/low_mean": 0.0012499999720603228,
|
| 2334 |
+
"clip_ratio/low_min": 0.0012499999720603228,
|
| 2335 |
+
"clip_ratio/region_mean": 0.0021253835438983515,
|
| 2336 |
+
"completions/clipped_ratio": 0.1171875,
|
| 2337 |
+
"completions/max_length": 51.0,
|
| 2338 |
+
"completions/mean_length": 15.8515625,
|
| 2339 |
+
"completions/min_length": 4.0,
|
| 2340 |
+
"epoch": 1.365650969529086,
|
| 2341 |
+
"grad_norm": 24.163527329310554,
|
| 2342 |
+
"kl": 24.496757203305606,
|
| 2343 |
+
"learning_rate": 5.000000000000002e-05,
|
| 2344 |
+
"loss": 0.23566541075706482,
|
| 2345 |
+
"memory(GiB)": 79.43,
|
| 2346 |
+
"reward": 0.3984375,
|
| 2347 |
+
"reward_std": 0.3253360986709595,
|
| 2348 |
+
"rewards/Response_no_think_reward_1/mean": 0.3984375,
|
| 2349 |
+
"rewards/Response_no_think_reward_1/std": 0.6912255883216858,
|
| 2350 |
+
"step": 123,
|
| 2351 |
+
"train_speed(iter/s)": 0.010181
|
| 2352 |
+
},
|
| 2353 |
+
{
|
| 2354 |
+
"clip_ratio/high_max": 0.0009191176941385493,
|
| 2355 |
+
"clip_ratio/high_mean": 0.0009191176941385493,
|
| 2356 |
+
"clip_ratio/low_mean": 0.025084137567318976,
|
| 2357 |
+
"clip_ratio/low_min": 0.025084137567318976,
|
| 2358 |
+
"clip_ratio/region_mean": 0.026003255392424762,
|
| 2359 |
+
"epoch": 1.3767313019390581,
|
| 2360 |
+
"grad_norm": 4.61798364138243,
|
| 2361 |
+
"kl": 5.482510100933723,
|
| 2362 |
+
"learning_rate": 4.841747534656763e-05,
|
| 2363 |
+
"loss": 0.03564080968499184,
|
| 2364 |
+
"memory(GiB)": 79.43,
|
| 2365 |
+
"step": 124,
|
| 2366 |
+
"train_speed(iter/s)": 0.010103
|
| 2367 |
+
},
|
| 2368 |
+
{
|
| 2369 |
+
"clip_ratio/high_max": 0.0004921259824186563,
|
| 2370 |
+
"clip_ratio/high_mean": 0.0004921259824186563,
|
| 2371 |
+
"clip_ratio/low_mean": 0.0009282178361900151,
|
| 2372 |
+
"clip_ratio/low_min": 0.0009282178361900151,
|
| 2373 |
+
"clip_ratio/region_mean": 0.0014203438186086714,
|
| 2374 |
+
"completions/clipped_ratio": 0.140625,
|
| 2375 |
+
"completions/max_length": 51.0,
|
| 2376 |
+
"completions/mean_length": 16.671875,
|
| 2377 |
+
"completions/min_length": 3.0,
|
| 2378 |
+
"epoch": 1.3878116343490305,
|
| 2379 |
+
"grad_norm": 1.7230796893481761,
|
| 2380 |
+
"kl": 1.7308420957997441,
|
| 2381 |
+
"learning_rate": 4.685236065835443e-05,
|
| 2382 |
+
"loss": 0.01634293608367443,
|
| 2383 |
+
"memory(GiB)": 79.43,
|
| 2384 |
+
"reward": 0.171875,
|
| 2385 |
+
"reward_std": 0.13258251547813416,
|
| 2386 |
+
"rewards/Response_no_think_reward_1/mean": 0.171875,
|
| 2387 |
+
"rewards/Response_no_think_reward_1/std": 0.8971465826034546,
|
| 2388 |
+
"step": 125,
|
| 2389 |
+
"train_speed(iter/s)": 0.009672
|
| 2390 |
+
},
|
| 2391 |
+
{
|
| 2392 |
+
"clip_ratio/high_max": 0.0003094059356953949,
|
| 2393 |
+
"clip_ratio/high_mean": 0.0003094059356953949,
|
| 2394 |
+
"clip_ratio/low_mean": 0.02542525064200163,
|
| 2395 |
+
"clip_ratio/low_min": 0.02542525064200163,
|
| 2396 |
+
"clip_ratio/region_mean": 0.025734656490385532,
|
| 2397 |
+
"epoch": 1.3988919667590027,
|
| 2398 |
+
"grad_norm": 0.6685767852950313,
|
| 2399 |
+
"kl": 2.058195663616061,
|
| 2400 |
+
"learning_rate": 4.530518418775733e-05,
|
| 2401 |
+
"loss": 0.008037411607801914,
|
| 2402 |
+
"memory(GiB)": 79.43,
|
| 2403 |
+
"step": 126,
|
| 2404 |
+
"train_speed(iter/s)": 0.009605
|
| 2405 |
+
},
|
| 2406 |
+
{
|
| 2407 |
+
"clip_ratio/high_max": 0.0035992932971566916,
|
| 2408 |
+
"clip_ratio/high_mean": 0.0035992932971566916,
|
| 2409 |
+
"clip_ratio/low_mean": 0.0026174120721407235,
|
| 2410 |
+
"clip_ratio/low_min": 0.0026174120721407235,
|
| 2411 |
+
"clip_ratio/region_mean": 0.006216705543920398,
|
| 2412 |
+
"completions/clipped_ratio": 0.078125,
|
| 2413 |
+
"completions/max_length": 51.0,
|
| 2414 |
+
"completions/mean_length": 13.1015625,
|
| 2415 |
+
"completions/min_length": 3.0,
|
| 2416 |
+
"epoch": 1.4099722991689752,
|
| 2417 |
+
"grad_norm": 0.898899557372063,
|
| 2418 |
+
"kl": 1.8244512832025066,
|
| 2419 |
+
"learning_rate": 4.3776468132724604e-05,
|
| 2420 |
+
"loss": 0.013885595835745335,
|
| 2421 |
+
"memory(GiB)": 79.43,
|
| 2422 |
+
"reward": 0.703125,
|
| 2423 |
+
"reward_std": 0.1841355264186859,
|
| 2424 |
+
"rewards/Response_no_think_reward_1/mean": 0.703125,
|
| 2425 |
+
"rewards/Response_no_think_reward_1/std": 0.552152156829834,
|
| 2426 |
+
"step": 127,
|
| 2427 |
+
"train_speed(iter/s)": 0.009212
|
| 2428 |
+
},
|
| 2429 |
+
{
|
| 2430 |
+
"clip_ratio/high_max": 0.00787301326636225,
|
| 2431 |
+
"clip_ratio/high_mean": 0.00787301326636225,
|
| 2432 |
+
"clip_ratio/low_mean": 0.003833794384263456,
|
| 2433 |
+
"clip_ratio/low_min": 0.003833794384263456,
|
| 2434 |
+
"clip_ratio/region_mean": 0.011706807999871671,
|
| 2435 |
+
"epoch": 1.4210526315789473,
|
| 2436 |
+
"grad_norm": 0.6061431338619399,
|
| 2437 |
+
"kl": 1.8278243900567759,
|
| 2438 |
+
"learning_rate": 4.2266728460505375e-05,
|
| 2439 |
+
"loss": 0.009479128755629063,
|
| 2440 |
+
"memory(GiB)": 79.43,
|
| 2441 |
+
"step": 128,
|
| 2442 |
+
"train_speed(iter/s)": 0.009157
|
| 2443 |
+
},
|
| 2444 |
+
{
|
| 2445 |
+
"clip_ratio/high_max": 0.0016208597226068377,
|
| 2446 |
+
"clip_ratio/high_mean": 0.0016208597226068377,
|
| 2447 |
+
"clip_ratio/low_mean": 0.004291799967177212,
|
| 2448 |
+
"clip_ratio/low_min": 0.004291799967177212,
|
| 2449 |
+
"clip_ratio/region_mean": 0.00591265968978405,
|
| 2450 |
+
"completions/clipped_ratio": 0.2421875,
|
| 2451 |
+
"completions/max_length": 51.0,
|
| 2452 |
+
"completions/mean_length": 13.8359375,
|
| 2453 |
+
"completions/min_length": 3.0,
|
| 2454 |
+
"epoch": 1.4321329639889195,
|
| 2455 |
+
"grad_norm": 4.256081630918958,
|
| 2456 |
+
"kl": 8.371784689603373,
|
| 2457 |
+
"learning_rate": 4.077647473350201e-05,
|
| 2458 |
+
"loss": 0.07109890133142471,
|
| 2459 |
+
"memory(GiB)": 79.43,
|
| 2460 |
+
"reward": 0.1640625,
|
| 2461 |
+
"reward_std": 0.12844271957874298,
|
| 2462 |
+
"rewards/Response_no_think_reward_1/mean": 0.1640625,
|
| 2463 |
+
"rewards/Response_no_think_reward_1/std": 0.9029901623725891,
|
| 2464 |
+
"step": 129,
|
| 2465 |
+
"train_speed(iter/s)": 0.008828
|
| 2466 |
+
},
|
| 2467 |
+
{
|
| 2468 |
+
"clip_ratio/high_max": 0.00940803368575871,
|
| 2469 |
+
"clip_ratio/high_mean": 0.00940803368575871,
|
| 2470 |
+
"clip_ratio/low_mean": 0.006534474319778383,
|
| 2471 |
+
"clip_ratio/low_min": 0.006534474319778383,
|
| 2472 |
+
"clip_ratio/region_mean": 0.01594250788912177,
|
| 2473 |
+
"epoch": 1.443213296398892,
|
| 2474 |
+
"grad_norm": 2.0975168570419678,
|
| 2475 |
+
"kl": 6.735093111405149,
|
| 2476 |
+
"learning_rate": 3.9306209937284346e-05,
|
| 2477 |
+
"loss": 0.05932926759123802,
|
| 2478 |
+
"memory(GiB)": 79.43,
|
| 2479 |
+
"step": 130,
|
| 2480 |
+
"train_speed(iter/s)": 0.008781
|
| 2481 |
+
}
|
| 2482 |
+
],
|
| 2483 |
+
"logging_steps": 1,
|
| 2484 |
+
"max_steps": 180,
|
| 2485 |
+
"num_input_tokens_seen": 0,
|
| 2486 |
+
"num_train_epochs": 2,
|
| 2487 |
+
"save_steps": 5,
|
| 2488 |
+
"stateful_callbacks": {
|
| 2489 |
+
"TrainerControl": {
|
| 2490 |
+
"args": {
|
| 2491 |
+
"should_epoch_stop": false,
|
| 2492 |
+
"should_evaluate": false,
|
| 2493 |
+
"should_log": false,
|
| 2494 |
+
"should_save": true,
|
| 2495 |
+
"should_training_stop": false
|
| 2496 |
+
},
|
| 2497 |
+
"attributes": {}
|
| 2498 |
+
}
|
| 2499 |
+
},
|
| 2500 |
+
"total_flos": 0.0,
|
| 2501 |
+
"train_batch_size": 8,
|
| 2502 |
+
"trial_name": null,
|
| 2503 |
+
"trial_params": null
|
| 2504 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc7d5aaa211d2aee94c15435720d80f37888d5443e6a3e6a917cafb907fde4f0
|
| 3 |
+
size 10168
|
zero_to_fp32.py
ADDED
|
@@ -0,0 +1,760 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python
|
| 2 |
+
|
| 3 |
+
# Copyright (c) Microsoft Corporation.
|
| 4 |
+
# SPDX-License-Identifier: Apache-2.0
|
| 5 |
+
|
| 6 |
+
# DeepSpeed Team
|
| 7 |
+
|
| 8 |
+
# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
|
| 9 |
+
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
|
| 10 |
+
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
|
| 11 |
+
# application.
|
| 12 |
+
#
|
| 13 |
+
# example:
|
| 14 |
+
# python zero_to_fp32.py . output_dir/
|
| 15 |
+
# or
|
| 16 |
+
# python zero_to_fp32.py . output_dir/ --safe_serialization
|
| 17 |
+
|
| 18 |
+
import argparse
|
| 19 |
+
import torch
|
| 20 |
+
import glob
|
| 21 |
+
import math
|
| 22 |
+
import os
|
| 23 |
+
import re
|
| 24 |
+
import gc
|
| 25 |
+
import json
|
| 26 |
+
import numpy as np
|
| 27 |
+
from tqdm import tqdm
|
| 28 |
+
from collections import OrderedDict
|
| 29 |
+
from dataclasses import dataclass
|
| 30 |
+
|
| 31 |
+
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
|
| 32 |
+
# DeepSpeed data structures it has to be available in the current python environment.
|
| 33 |
+
from deepspeed.utils import logger
|
| 34 |
+
from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
|
| 35 |
+
FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
|
| 36 |
+
FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
@dataclass
|
| 40 |
+
class zero_model_state:
|
| 41 |
+
buffers: dict()
|
| 42 |
+
param_shapes: dict()
|
| 43 |
+
shared_params: list
|
| 44 |
+
ds_version: int
|
| 45 |
+
frozen_param_shapes: dict()
|
| 46 |
+
frozen_param_fragments: dict()
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
debug = 0
|
| 50 |
+
|
| 51 |
+
# load to cpu
|
| 52 |
+
device = torch.device('cpu')
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
def atoi(text):
|
| 56 |
+
return int(text) if text.isdigit() else text
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
def natural_keys(text):
|
| 60 |
+
'''
|
| 61 |
+
alist.sort(key=natural_keys) sorts in human order
|
| 62 |
+
http://nedbatchelder.com/blog/200712/human_sorting.html
|
| 63 |
+
(See Toothy's implementation in the comments)
|
| 64 |
+
'''
|
| 65 |
+
return [atoi(c) for c in re.split(r'(\d+)', text)]
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
def get_model_state_file(checkpoint_dir, zero_stage):
|
| 69 |
+
if not os.path.isdir(checkpoint_dir):
|
| 70 |
+
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
|
| 71 |
+
|
| 72 |
+
# there should be only one file
|
| 73 |
+
if zero_stage <= 2:
|
| 74 |
+
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
|
| 75 |
+
elif zero_stage == 3:
|
| 76 |
+
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
|
| 77 |
+
|
| 78 |
+
if not os.path.exists(file):
|
| 79 |
+
raise FileNotFoundError(f"can't find model states file at '{file}'")
|
| 80 |
+
|
| 81 |
+
return file
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
def get_checkpoint_files(checkpoint_dir, glob_pattern):
|
| 85 |
+
# XXX: need to test that this simple glob rule works for multi-node setup too
|
| 86 |
+
ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
|
| 87 |
+
|
| 88 |
+
if len(ckpt_files) == 0:
|
| 89 |
+
raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
|
| 90 |
+
|
| 91 |
+
return ckpt_files
|
| 92 |
+
|
| 93 |
+
|
| 94 |
+
def get_optim_files(checkpoint_dir):
|
| 95 |
+
return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
def get_model_state_files(checkpoint_dir):
|
| 99 |
+
return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
|
| 100 |
+
|
| 101 |
+
|
| 102 |
+
def parse_model_states(files):
|
| 103 |
+
zero_model_states = []
|
| 104 |
+
for file in files:
|
| 105 |
+
state_dict = torch.load(file, map_location=device, weights_only=False)
|
| 106 |
+
|
| 107 |
+
if BUFFER_NAMES not in state_dict:
|
| 108 |
+
raise ValueError(f"{file} is not a model state checkpoint")
|
| 109 |
+
buffer_names = state_dict[BUFFER_NAMES]
|
| 110 |
+
if debug:
|
| 111 |
+
print("Found buffers:", buffer_names)
|
| 112 |
+
|
| 113 |
+
# recover just the buffers while restoring them to fp32 if they were saved in fp16
|
| 114 |
+
buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
|
| 115 |
+
param_shapes = state_dict[PARAM_SHAPES]
|
| 116 |
+
|
| 117 |
+
# collect parameters that are included in param_shapes
|
| 118 |
+
param_names = []
|
| 119 |
+
for s in param_shapes:
|
| 120 |
+
for name in s.keys():
|
| 121 |
+
param_names.append(name)
|
| 122 |
+
|
| 123 |
+
# update with frozen parameters
|
| 124 |
+
frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
|
| 125 |
+
if frozen_param_shapes is not None:
|
| 126 |
+
if debug:
|
| 127 |
+
print(f"Found frozen_param_shapes: {frozen_param_shapes}")
|
| 128 |
+
param_names += list(frozen_param_shapes.keys())
|
| 129 |
+
|
| 130 |
+
# handle shared params
|
| 131 |
+
shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
|
| 132 |
+
|
| 133 |
+
ds_version = state_dict.get(DS_VERSION, None)
|
| 134 |
+
|
| 135 |
+
frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
|
| 136 |
+
|
| 137 |
+
z_model_state = zero_model_state(buffers=buffers,
|
| 138 |
+
param_shapes=param_shapes,
|
| 139 |
+
shared_params=shared_params,
|
| 140 |
+
ds_version=ds_version,
|
| 141 |
+
frozen_param_shapes=frozen_param_shapes,
|
| 142 |
+
frozen_param_fragments=frozen_param_fragments)
|
| 143 |
+
zero_model_states.append(z_model_state)
|
| 144 |
+
|
| 145 |
+
return zero_model_states
|
| 146 |
+
|
| 147 |
+
|
| 148 |
+
def parse_optim_states(files, ds_checkpoint_dir):
|
| 149 |
+
total_files = len(files)
|
| 150 |
+
state_dicts = []
|
| 151 |
+
for f in tqdm(files, desc='Loading checkpoint shards'):
|
| 152 |
+
state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
|
| 153 |
+
# immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
|
| 154 |
+
# and also handle the case where it was already removed by another helper script
|
| 155 |
+
state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
|
| 156 |
+
state_dicts.append(state_dict)
|
| 157 |
+
|
| 158 |
+
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
|
| 159 |
+
raise ValueError(f"{files[0]} is not a zero checkpoint")
|
| 160 |
+
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
|
| 161 |
+
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
|
| 162 |
+
|
| 163 |
+
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
|
| 164 |
+
# parameters can be different from data parallelism for non-expert parameters. So we can just
|
| 165 |
+
# use the max of the partition_count to get the dp world_size.
|
| 166 |
+
|
| 167 |
+
if type(world_size) is list:
|
| 168 |
+
world_size = max(world_size)
|
| 169 |
+
|
| 170 |
+
if world_size != total_files:
|
| 171 |
+
raise ValueError(
|
| 172 |
+
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
|
| 173 |
+
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
|
| 174 |
+
)
|
| 175 |
+
|
| 176 |
+
# the groups are named differently in each stage
|
| 177 |
+
if zero_stage <= 2:
|
| 178 |
+
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
|
| 179 |
+
elif zero_stage == 3:
|
| 180 |
+
fp32_groups_key = FP32_FLAT_GROUPS
|
| 181 |
+
else:
|
| 182 |
+
raise ValueError(f"unknown zero stage {zero_stage}")
|
| 183 |
+
|
| 184 |
+
fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
|
| 185 |
+
return zero_stage, world_size, fp32_flat_groups
|
| 186 |
+
|
| 187 |
+
|
| 188 |
+
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
|
| 189 |
+
"""
|
| 190 |
+
Returns fp32 state_dict reconstructed from ds checkpoint
|
| 191 |
+
|
| 192 |
+
Args:
|
| 193 |
+
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
|
| 194 |
+
|
| 195 |
+
"""
|
| 196 |
+
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
|
| 197 |
+
|
| 198 |
+
optim_files = get_optim_files(ds_checkpoint_dir)
|
| 199 |
+
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
|
| 200 |
+
print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
|
| 201 |
+
|
| 202 |
+
model_files = get_model_state_files(ds_checkpoint_dir)
|
| 203 |
+
|
| 204 |
+
zero_model_states = parse_model_states(model_files)
|
| 205 |
+
print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
|
| 206 |
+
|
| 207 |
+
if zero_stage <= 2:
|
| 208 |
+
return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
| 209 |
+
exclude_frozen_parameters)
|
| 210 |
+
elif zero_stage == 3:
|
| 211 |
+
return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
| 212 |
+
exclude_frozen_parameters)
|
| 213 |
+
|
| 214 |
+
|
| 215 |
+
def _zero2_merge_frozen_params(state_dict, zero_model_states):
|
| 216 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
| 217 |
+
return
|
| 218 |
+
|
| 219 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
| 220 |
+
frozen_param_fragments = zero_model_states[0].frozen_param_fragments
|
| 221 |
+
|
| 222 |
+
if debug:
|
| 223 |
+
num_elem = sum(s.numel() for s in frozen_param_shapes.values())
|
| 224 |
+
print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
| 225 |
+
|
| 226 |
+
wanted_params = len(frozen_param_shapes)
|
| 227 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
| 228 |
+
avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
|
| 229 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
| 230 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
| 231 |
+
|
| 232 |
+
total_params = 0
|
| 233 |
+
total_numel = 0
|
| 234 |
+
for name, shape in frozen_param_shapes.items():
|
| 235 |
+
total_params += 1
|
| 236 |
+
unpartitioned_numel = shape.numel()
|
| 237 |
+
total_numel += unpartitioned_numel
|
| 238 |
+
|
| 239 |
+
state_dict[name] = frozen_param_fragments[name]
|
| 240 |
+
|
| 241 |
+
if debug:
|
| 242 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
| 243 |
+
|
| 244 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
| 245 |
+
|
| 246 |
+
|
| 247 |
+
def _has_callable(obj, fn):
|
| 248 |
+
attr = getattr(obj, fn, None)
|
| 249 |
+
return callable(attr)
|
| 250 |
+
|
| 251 |
+
|
| 252 |
+
def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
| 253 |
+
param_shapes = zero_model_states[0].param_shapes
|
| 254 |
+
|
| 255 |
+
# Reconstruction protocol:
|
| 256 |
+
#
|
| 257 |
+
# XXX: document this
|
| 258 |
+
|
| 259 |
+
if debug:
|
| 260 |
+
for i in range(world_size):
|
| 261 |
+
for j in range(len(fp32_flat_groups[0])):
|
| 262 |
+
print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
|
| 263 |
+
|
| 264 |
+
# XXX: memory usage doubles here (zero2)
|
| 265 |
+
num_param_groups = len(fp32_flat_groups[0])
|
| 266 |
+
merged_single_partition_of_fp32_groups = []
|
| 267 |
+
for i in range(num_param_groups):
|
| 268 |
+
merged_partitions = [sd[i] for sd in fp32_flat_groups]
|
| 269 |
+
full_single_fp32_vector = torch.cat(merged_partitions, 0)
|
| 270 |
+
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
|
| 271 |
+
avail_numel = sum(
|
| 272 |
+
[full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
|
| 273 |
+
|
| 274 |
+
if debug:
|
| 275 |
+
wanted_params = sum([len(shapes) for shapes in param_shapes])
|
| 276 |
+
wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
|
| 277 |
+
# not asserting if there is a mismatch due to possible padding
|
| 278 |
+
print(f"Have {avail_numel} numels to process.")
|
| 279 |
+
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
| 280 |
+
|
| 281 |
+
# params
|
| 282 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
| 283 |
+
# out-of-core computing solution
|
| 284 |
+
total_numel = 0
|
| 285 |
+
total_params = 0
|
| 286 |
+
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
|
| 287 |
+
offset = 0
|
| 288 |
+
avail_numel = full_single_fp32_vector.numel()
|
| 289 |
+
for name, shape in shapes.items():
|
| 290 |
+
|
| 291 |
+
unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
|
| 292 |
+
total_numel += unpartitioned_numel
|
| 293 |
+
total_params += 1
|
| 294 |
+
|
| 295 |
+
if debug:
|
| 296 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
| 297 |
+
state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
|
| 298 |
+
offset += unpartitioned_numel
|
| 299 |
+
|
| 300 |
+
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
|
| 301 |
+
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
|
| 302 |
+
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
|
| 303 |
+
# live optimizer object, so we are checking that the numbers are within the right range
|
| 304 |
+
align_to = 2 * world_size
|
| 305 |
+
|
| 306 |
+
def zero2_align(x):
|
| 307 |
+
return align_to * math.ceil(x / align_to)
|
| 308 |
+
|
| 309 |
+
if debug:
|
| 310 |
+
print(f"original offset={offset}, avail_numel={avail_numel}")
|
| 311 |
+
|
| 312 |
+
offset = zero2_align(offset)
|
| 313 |
+
avail_numel = zero2_align(avail_numel)
|
| 314 |
+
|
| 315 |
+
if debug:
|
| 316 |
+
print(f"aligned offset={offset}, avail_numel={avail_numel}")
|
| 317 |
+
|
| 318 |
+
# Sanity check
|
| 319 |
+
if offset != avail_numel:
|
| 320 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
| 321 |
+
|
| 322 |
+
print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
|
| 323 |
+
|
| 324 |
+
|
| 325 |
+
def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
| 326 |
+
exclude_frozen_parameters):
|
| 327 |
+
state_dict = OrderedDict()
|
| 328 |
+
|
| 329 |
+
# buffers
|
| 330 |
+
buffers = zero_model_states[0].buffers
|
| 331 |
+
state_dict.update(buffers)
|
| 332 |
+
if debug:
|
| 333 |
+
print(f"added {len(buffers)} buffers")
|
| 334 |
+
|
| 335 |
+
if not exclude_frozen_parameters:
|
| 336 |
+
_zero2_merge_frozen_params(state_dict, zero_model_states)
|
| 337 |
+
|
| 338 |
+
_zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
| 339 |
+
|
| 340 |
+
# recover shared parameters
|
| 341 |
+
for pair in zero_model_states[0].shared_params:
|
| 342 |
+
if pair[1] in state_dict:
|
| 343 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
| 344 |
+
|
| 345 |
+
return state_dict
|
| 346 |
+
|
| 347 |
+
|
| 348 |
+
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
|
| 349 |
+
remainder = unpartitioned_numel % world_size
|
| 350 |
+
padding_numel = (world_size - remainder) if remainder else 0
|
| 351 |
+
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
|
| 352 |
+
return partitioned_numel, padding_numel
|
| 353 |
+
|
| 354 |
+
|
| 355 |
+
def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
|
| 356 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
| 357 |
+
return
|
| 358 |
+
|
| 359 |
+
if debug:
|
| 360 |
+
for i in range(world_size):
|
| 361 |
+
num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
|
| 362 |
+
print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
| 363 |
+
|
| 364 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
| 365 |
+
wanted_params = len(frozen_param_shapes)
|
| 366 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
| 367 |
+
avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
|
| 368 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
| 369 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
| 370 |
+
|
| 371 |
+
total_params = 0
|
| 372 |
+
total_numel = 0
|
| 373 |
+
for name, shape in zero_model_states[0].frozen_param_shapes.items():
|
| 374 |
+
total_params += 1
|
| 375 |
+
unpartitioned_numel = shape.numel()
|
| 376 |
+
total_numel += unpartitioned_numel
|
| 377 |
+
|
| 378 |
+
param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
|
| 379 |
+
state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
|
| 380 |
+
|
| 381 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
| 382 |
+
|
| 383 |
+
if debug:
|
| 384 |
+
print(
|
| 385 |
+
f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
| 386 |
+
)
|
| 387 |
+
|
| 388 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
| 389 |
+
|
| 390 |
+
|
| 391 |
+
class GatheredTensor:
|
| 392 |
+
"""
|
| 393 |
+
A pseudo tensor that collects partitioned weights.
|
| 394 |
+
It is more memory efficient when there are multiple groups.
|
| 395 |
+
"""
|
| 396 |
+
|
| 397 |
+
def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
|
| 398 |
+
self.flat_groups = flat_groups
|
| 399 |
+
self.flat_groups_offset = flat_groups_offset
|
| 400 |
+
self.offset = offset
|
| 401 |
+
self.partitioned_numel = partitioned_numel
|
| 402 |
+
self.shape = shape
|
| 403 |
+
self.dtype = self.flat_groups[0][0].dtype
|
| 404 |
+
|
| 405 |
+
def contiguous(self):
|
| 406 |
+
"""
|
| 407 |
+
Merge partitioned weights from flat_groups into a single tensor.
|
| 408 |
+
"""
|
| 409 |
+
end_idx = self.offset + self.partitioned_numel
|
| 410 |
+
world_size = len(self.flat_groups)
|
| 411 |
+
pad_flat_param_chunks = []
|
| 412 |
+
|
| 413 |
+
for rank_i in range(world_size):
|
| 414 |
+
# for each rank, we need to collect weights from related group/groups
|
| 415 |
+
flat_groups_at_rank_i = self.flat_groups[rank_i]
|
| 416 |
+
start_group_id = None
|
| 417 |
+
end_group_id = None
|
| 418 |
+
for group_id in range(len(self.flat_groups_offset)):
|
| 419 |
+
if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
|
| 420 |
+
start_group_id = group_id
|
| 421 |
+
if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
|
| 422 |
+
end_group_id = group_id
|
| 423 |
+
break
|
| 424 |
+
# collect weights from related group/groups
|
| 425 |
+
for group_id in range(start_group_id, end_group_id + 1):
|
| 426 |
+
flat_tensor = flat_groups_at_rank_i[group_id]
|
| 427 |
+
start_offset = self.offset - self.flat_groups_offset[group_id]
|
| 428 |
+
end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
|
| 429 |
+
pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
|
| 430 |
+
|
| 431 |
+
# collect weights from all ranks
|
| 432 |
+
pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
|
| 433 |
+
param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
|
| 434 |
+
return param
|
| 435 |
+
|
| 436 |
+
|
| 437 |
+
def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
| 438 |
+
param_shapes = zero_model_states[0].param_shapes
|
| 439 |
+
avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
|
| 440 |
+
|
| 441 |
+
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
|
| 442 |
+
# param, re-consolidating each param, while dealing with padding if any
|
| 443 |
+
|
| 444 |
+
# merge list of dicts, preserving order
|
| 445 |
+
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
|
| 446 |
+
|
| 447 |
+
if debug:
|
| 448 |
+
for i in range(world_size):
|
| 449 |
+
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
|
| 450 |
+
|
| 451 |
+
wanted_params = len(param_shapes)
|
| 452 |
+
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
|
| 453 |
+
# not asserting if there is a mismatch due to possible padding
|
| 454 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
| 455 |
+
print(f"Trainable params: Have {avail_numel} numels to process.")
|
| 456 |
+
print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
|
| 457 |
+
|
| 458 |
+
# params
|
| 459 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
| 460 |
+
# out-of-core computing solution
|
| 461 |
+
offset = 0
|
| 462 |
+
total_numel = 0
|
| 463 |
+
total_params = 0
|
| 464 |
+
flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
|
| 465 |
+
for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
|
| 466 |
+
unpartitioned_numel = shape.numel()
|
| 467 |
+
total_numel += unpartitioned_numel
|
| 468 |
+
total_params += 1
|
| 469 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
| 470 |
+
|
| 471 |
+
if debug:
|
| 472 |
+
print(
|
| 473 |
+
f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
| 474 |
+
)
|
| 475 |
+
|
| 476 |
+
# memory efficient tensor
|
| 477 |
+
tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
|
| 478 |
+
state_dict[name] = tensor
|
| 479 |
+
offset += partitioned_numel
|
| 480 |
+
|
| 481 |
+
offset *= world_size
|
| 482 |
+
|
| 483 |
+
# Sanity check
|
| 484 |
+
if offset != avail_numel:
|
| 485 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
| 486 |
+
|
| 487 |
+
print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
|
| 488 |
+
|
| 489 |
+
|
| 490 |
+
def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
| 491 |
+
exclude_frozen_parameters):
|
| 492 |
+
state_dict = OrderedDict()
|
| 493 |
+
|
| 494 |
+
# buffers
|
| 495 |
+
buffers = zero_model_states[0].buffers
|
| 496 |
+
state_dict.update(buffers)
|
| 497 |
+
if debug:
|
| 498 |
+
print(f"added {len(buffers)} buffers")
|
| 499 |
+
|
| 500 |
+
if not exclude_frozen_parameters:
|
| 501 |
+
_zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
|
| 502 |
+
|
| 503 |
+
_zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
| 504 |
+
|
| 505 |
+
# recover shared parameters
|
| 506 |
+
for pair in zero_model_states[0].shared_params:
|
| 507 |
+
if pair[1] in state_dict:
|
| 508 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
| 509 |
+
|
| 510 |
+
return state_dict
|
| 511 |
+
|
| 512 |
+
|
| 513 |
+
def to_torch_tensor(state_dict, return_empty_tensor=False):
|
| 514 |
+
"""
|
| 515 |
+
Convert state_dict of GatheredTensor to torch tensor
|
| 516 |
+
"""
|
| 517 |
+
torch_state_dict = {}
|
| 518 |
+
converted_tensors = {}
|
| 519 |
+
for name, tensor in state_dict.items():
|
| 520 |
+
tensor_id = id(tensor)
|
| 521 |
+
if tensor_id in converted_tensors: # shared tensors
|
| 522 |
+
shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
|
| 523 |
+
torch_state_dict[name] = shared_tensor
|
| 524 |
+
else:
|
| 525 |
+
converted_tensors[tensor_id] = name
|
| 526 |
+
if return_empty_tensor:
|
| 527 |
+
torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
|
| 528 |
+
else:
|
| 529 |
+
torch_state_dict[name] = tensor.contiguous()
|
| 530 |
+
return torch_state_dict
|
| 531 |
+
|
| 532 |
+
|
| 533 |
+
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
|
| 534 |
+
tag=None,
|
| 535 |
+
exclude_frozen_parameters=False,
|
| 536 |
+
lazy_mode=False):
|
| 537 |
+
"""
|
| 538 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
|
| 539 |
+
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
|
| 540 |
+
via a model hub.
|
| 541 |
+
|
| 542 |
+
Args:
|
| 543 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder
|
| 544 |
+
- ``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``
|
| 545 |
+
- ``exclude_frozen_parameters``: exclude frozen parameters
|
| 546 |
+
- ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
|
| 547 |
+
Convert the pesduo tensor to torch tensor by ``.contiguous()``
|
| 548 |
+
|
| 549 |
+
Returns:
|
| 550 |
+
- pytorch ``state_dict``
|
| 551 |
+
|
| 552 |
+
A typical usage might be ::
|
| 553 |
+
|
| 554 |
+
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
| 555 |
+
# do the training and checkpoint saving
|
| 556 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
|
| 557 |
+
model = model.cpu() # move to cpu
|
| 558 |
+
model.load_state_dict(state_dict)
|
| 559 |
+
# submit to model hub or save the model to share with others
|
| 560 |
+
|
| 561 |
+
In this example the ``model`` will no longer be usable in the deepspeed context of the same
|
| 562 |
+
application. i.e. you will need to re-initialize the deepspeed engine, since
|
| 563 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
| 564 |
+
|
| 565 |
+
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
|
| 566 |
+
|
| 567 |
+
Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
|
| 568 |
+
You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
|
| 569 |
+
the checkpoint. Or you can load state_dict in lazy mode ::
|
| 570 |
+
|
| 571 |
+
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
| 572 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
|
| 573 |
+
for name, lazy_tensor in state_dict.item():
|
| 574 |
+
tensor = lazy_tensor.contiguous() # to cpu
|
| 575 |
+
print(name, tensor)
|
| 576 |
+
# del tensor to release memory if it no longer in use
|
| 577 |
+
"""
|
| 578 |
+
if tag is None:
|
| 579 |
+
latest_path = os.path.join(checkpoint_dir, 'latest')
|
| 580 |
+
if os.path.isfile(latest_path):
|
| 581 |
+
with open(latest_path, 'r') as fd:
|
| 582 |
+
tag = fd.read().strip()
|
| 583 |
+
else:
|
| 584 |
+
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
|
| 585 |
+
|
| 586 |
+
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
|
| 587 |
+
|
| 588 |
+
if not os.path.isdir(ds_checkpoint_dir):
|
| 589 |
+
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
|
| 590 |
+
|
| 591 |
+
state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
|
| 592 |
+
if lazy_mode:
|
| 593 |
+
return state_dict
|
| 594 |
+
else:
|
| 595 |
+
return to_torch_tensor(state_dict)
|
| 596 |
+
|
| 597 |
+
|
| 598 |
+
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
|
| 599 |
+
output_dir,
|
| 600 |
+
max_shard_size="5GB",
|
| 601 |
+
safe_serialization=False,
|
| 602 |
+
tag=None,
|
| 603 |
+
exclude_frozen_parameters=False):
|
| 604 |
+
"""
|
| 605 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
|
| 606 |
+
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
|
| 607 |
+
|
| 608 |
+
Args:
|
| 609 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
| 610 |
+
- ``output_dir``: directory to the pytorch fp32 state_dict output files
|
| 611 |
+
- ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
|
| 612 |
+
- ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
|
| 613 |
+
- ``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``
|
| 614 |
+
- ``exclude_frozen_parameters``: exclude frozen parameters
|
| 615 |
+
"""
|
| 616 |
+
|
| 617 |
+
# Dependency pre-check
|
| 618 |
+
if safe_serialization:
|
| 619 |
+
try:
|
| 620 |
+
from safetensors.torch import save_file
|
| 621 |
+
except ImportError:
|
| 622 |
+
print('If you want to use `safe_serialization`, please `pip install safetensors`')
|
| 623 |
+
raise
|
| 624 |
+
if max_shard_size is not None:
|
| 625 |
+
try:
|
| 626 |
+
from huggingface_hub import split_torch_state_dict_into_shards
|
| 627 |
+
except ImportError:
|
| 628 |
+
print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
|
| 629 |
+
raise
|
| 630 |
+
|
| 631 |
+
# Convert zero checkpoint to state_dict
|
| 632 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
|
| 633 |
+
tag,
|
| 634 |
+
exclude_frozen_parameters,
|
| 635 |
+
lazy_mode=True)
|
| 636 |
+
|
| 637 |
+
# Shard the model if it is too big.
|
| 638 |
+
weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
|
| 639 |
+
if max_shard_size is not None:
|
| 640 |
+
filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
|
| 641 |
+
# an memory-efficient approach for sharding
|
| 642 |
+
empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
|
| 643 |
+
state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
|
| 644 |
+
filename_pattern=filename_pattern,
|
| 645 |
+
max_shard_size=max_shard_size)
|
| 646 |
+
else:
|
| 647 |
+
from collections import namedtuple
|
| 648 |
+
StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
|
| 649 |
+
state_dict_split = StateDictSplit(is_sharded=False,
|
| 650 |
+
filename_to_tensors={weights_name: list(state_dict.keys())})
|
| 651 |
+
|
| 652 |
+
# Save the model by shard
|
| 653 |
+
os.makedirs(output_dir, exist_ok=True)
|
| 654 |
+
filename_to_tensors = state_dict_split.filename_to_tensors.items()
|
| 655 |
+
for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
|
| 656 |
+
shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
|
| 657 |
+
shard_state_dict = to_torch_tensor(shard_state_dict)
|
| 658 |
+
output_path = os.path.join(output_dir, shard_file)
|
| 659 |
+
if safe_serialization:
|
| 660 |
+
save_file(shard_state_dict, output_path, metadata={"format": "pt"})
|
| 661 |
+
else:
|
| 662 |
+
torch.save(shard_state_dict, output_path)
|
| 663 |
+
# release the memory of current shard
|
| 664 |
+
for tensor_name in list(shard_state_dict.keys()):
|
| 665 |
+
del state_dict[tensor_name]
|
| 666 |
+
del shard_state_dict[tensor_name]
|
| 667 |
+
del shard_state_dict
|
| 668 |
+
gc.collect()
|
| 669 |
+
|
| 670 |
+
# Save index if sharded
|
| 671 |
+
if state_dict_split.is_sharded:
|
| 672 |
+
index = {
|
| 673 |
+
"metadata": state_dict_split.metadata,
|
| 674 |
+
"weight_map": state_dict_split.tensor_to_filename,
|
| 675 |
+
}
|
| 676 |
+
save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
|
| 677 |
+
save_index_file = os.path.join(output_dir, save_index_file)
|
| 678 |
+
with open(save_index_file, "w", encoding="utf-8") as f:
|
| 679 |
+
content = json.dumps(index, indent=2, sort_keys=True) + "\n"
|
| 680 |
+
f.write(content)
|
| 681 |
+
|
| 682 |
+
|
| 683 |
+
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
|
| 684 |
+
"""
|
| 685 |
+
1. Put the provided model to cpu
|
| 686 |
+
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
|
| 687 |
+
3. Load it into the provided model
|
| 688 |
+
|
| 689 |
+
Args:
|
| 690 |
+
- ``model``: the model object to update
|
| 691 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
| 692 |
+
- ``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``
|
| 693 |
+
|
| 694 |
+
Returns:
|
| 695 |
+
- ``model`: modified model
|
| 696 |
+
|
| 697 |
+
Make sure you have plenty of CPU memory available before you call this function. If you don't
|
| 698 |
+
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
|
| 699 |
+
conveniently placed for you in the checkpoint folder.
|
| 700 |
+
|
| 701 |
+
A typical usage might be ::
|
| 702 |
+
|
| 703 |
+
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
|
| 704 |
+
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
|
| 705 |
+
# submit to model hub or save the model to share with others
|
| 706 |
+
|
| 707 |
+
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
|
| 708 |
+
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
|
| 709 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
| 710 |
+
|
| 711 |
+
"""
|
| 712 |
+
logger.info(f"Extracting fp32 weights")
|
| 713 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
| 714 |
+
|
| 715 |
+
logger.info(f"Overwriting model with fp32 weights")
|
| 716 |
+
model = model.cpu()
|
| 717 |
+
model.load_state_dict(state_dict, strict=False)
|
| 718 |
+
|
| 719 |
+
return model
|
| 720 |
+
|
| 721 |
+
|
| 722 |
+
if __name__ == "__main__":
|
| 723 |
+
parser = argparse.ArgumentParser()
|
| 724 |
+
parser.add_argument("checkpoint_dir",
|
| 725 |
+
type=str,
|
| 726 |
+
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
|
| 727 |
+
parser.add_argument("output_dir",
|
| 728 |
+
type=str,
|
| 729 |
+
help="directory to the pytorch fp32 state_dict output files"
|
| 730 |
+
"(e.g. path/checkpoint-12-output/)")
|
| 731 |
+
parser.add_argument(
|
| 732 |
+
"--max_shard_size",
|
| 733 |
+
type=str,
|
| 734 |
+
default="5GB",
|
| 735 |
+
help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
|
| 736 |
+
"lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
|
| 737 |
+
"We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
|
| 738 |
+
"without CPU OOM issues.")
|
| 739 |
+
parser.add_argument(
|
| 740 |
+
"--safe_serialization",
|
| 741 |
+
default=False,
|
| 742 |
+
action='store_true',
|
| 743 |
+
help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
|
| 744 |
+
parser.add_argument("-t",
|
| 745 |
+
"--tag",
|
| 746 |
+
type=str,
|
| 747 |
+
default=None,
|
| 748 |
+
help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
|
| 749 |
+
parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
|
| 750 |
+
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
|
| 751 |
+
args = parser.parse_args()
|
| 752 |
+
|
| 753 |
+
debug = args.debug
|
| 754 |
+
|
| 755 |
+
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
|
| 756 |
+
args.output_dir,
|
| 757 |
+
max_shard_size=args.max_shard_size,
|
| 758 |
+
safe_serialization=args.safe_serialization,
|
| 759 |
+
tag=args.tag,
|
| 760 |
+
exclude_frozen_parameters=args.exclude_frozen_parameters)
|