Training in progress, step 208, checkpoint

checkpoint-208/README.md

@@ -0,0 +1,202 @@
+---
+base_model: mistralai/Mistral-Large-Instruct-2411
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- 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. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+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).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.2

checkpoint-208/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "mistralai/Mistral-Large-Instruct-2411",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [
+    "embed_tokens",
+    "lm_head"
+  ],
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-208/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd29413c0c56980b112ca5f63fe6d184c960858904e68755a10ae321ff017d5f
+size 6086108064

checkpoint-208/global_step208/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccf8027cd751ad1e7ef7088dc00a3bbe35dea9cd98f7d406b845f374dc5a9ac2
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8359e904479f454ede8aba8167f0c486ba0135f3cb031cd3f670c19840b0745
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3436f93d9ba364e98c71f831307cb7b039e490df88a784dd9158ada1fff8664
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c58405a8e1b0f1426237197ba06bd256eaf3ca6340035327cffe09823e7f2de7
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e8d2c59af4b28db557cc422b1d9000045e67d93089a028152bda937c52053a3
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d66c34ab4d0536a53df0f637689ada4d705004c6cb05da41c66077ece3e923d5
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8160ad94ea61c6af40a87222e8ac299eb794334f2900849fc22b6a809487d609
+size 2682523907

checkpoint-208/global_step208/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c17e3f86b9a6c9d5e17ebf661e5f7386f22728902bd7fe21ac9ecf18017340a2
+size 2682523907

checkpoint-208/global_step208/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68030339bb95b80cdd1bab9f88f440c3032b9a8217bb7e883b60394f41166fa0
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95c1c5e14f17a71bc342b14a51148d97ed6955bf93824204950870a47a794faa
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8cb13952fc543bd05b5efae23600d4a0fbdf5024b0fe7da3cd6932bff40c094
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bad4207568c63337a4894a9475721143841ccd301d6e6e48e633d35ebc4a1904
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d740532c4e226309015c2b01536bec17da378b578e649467172e215d0bd330
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bebe2459e4e00f61771479d0e489bb9363a584fbf5d4bec498326fc384f24ceb
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4578ee208c672f3c0bedfc9388e5be026d7278c8415913ec677be0eb01c89d5
+size 1226797

checkpoint-208/global_step208/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0166902d97aae4de51c361bf001e310c45529217484219b14c35f06c02a830ee
+size 1226797

checkpoint-208/latest

@@ -0,0 +1 @@
+global_step208

checkpoint-208/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7daed31adb2bcbcd1158f9db2ccdd264af0ced94bdb99feb4e9ef0e279276c1
+size 16389

checkpoint-208/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7c5bd4942333ed5ac4884ef843e90f977dfbc43908dd77c365be358cc2c380d
+size 16389

checkpoint-208/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad8a805fbd04678f973acc547c92a2c6d0dfac9a5d4a377664252829901975fc
+size 16389

checkpoint-208/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abe53b5e80cf268a159920ac012c4533ca6756a46fa0610476614716081578ec
+size 16389

checkpoint-208/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e6315f475abf1927af6548bedd63a782d906c72558e973de1ad0a9d87d2654b
+size 16389

checkpoint-208/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14bff758a59f25da387a780b3f72f52f99a5a3544ec25495045a4b88134c6699
+size 16389

checkpoint-208/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f969e72443075d791a8a5569d91ebaf40495d7c45d6d4bb9af1ad2d431bdfa0b
+size 16389

checkpoint-208/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f45947f8980f04bca549a5a7e70d5cdc7d0e099fe6fee06d9688a8e71343b1d
+size 16389

checkpoint-208/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06f96c7078f551e3c505d74211b9fddebdf5d4d89238dcb0077c545a2f87302f
+size 1401

checkpoint-208/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-208/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b968b8dc352f42192367337c78ccc61e1eaddc6d641a579372d4f20694beb7a
+size 587562

checkpoint-208/trainer_state.json

@@ -0,0 +1,1490 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.49939975990396157,
+  "eval_steps": 500,
+  "global_step": 208,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0024009603841536613,
+      "grad_norm": 3.307639305281489,
+      "learning_rate": 0.0,
+      "loss": 1.027,
+      "step": 1
+    },
+    {
+      "epoch": 0.004801920768307323,
+      "grad_norm": 3.4815729475698585,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 1.0887,
+      "step": 2
+    },
+    {
+      "epoch": 0.007202881152460984,
+      "grad_norm": 3.1919511484422025,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 0.9992,
+      "step": 3
+    },
+    {
+      "epoch": 0.009603841536614645,
+      "grad_norm": 3.192234811391807,
+      "learning_rate": 7.5e-07,
+      "loss": 1.0676,
+      "step": 4
+    },
+    {
+      "epoch": 0.012004801920768308,
+      "grad_norm": 3.147628945570283,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.0932,
+      "step": 5
+    },
+    {
+      "epoch": 0.014405762304921969,
+      "grad_norm": 2.9173656342980085,
+      "learning_rate": 1.25e-06,
+      "loss": 1.0108,
+      "step": 6
+    },
+    {
+      "epoch": 0.01680672268907563,
+      "grad_norm": 2.965584304042035,
+      "learning_rate": 1.5e-06,
+      "loss": 1.0785,
+      "step": 7
+    },
+    {
+      "epoch": 0.01920768307322929,
+      "grad_norm": 2.6367740096927643,
+      "learning_rate": 1.75e-06,
+      "loss": 1.0562,
+      "step": 8
+    },
+    {
+      "epoch": 0.021608643457382955,
+      "grad_norm": 2.468163591370129,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.0243,
+      "step": 9
+    },
+    {
+      "epoch": 0.024009603841536616,
+      "grad_norm": 2.372272215735556,
+      "learning_rate": 2.25e-06,
+      "loss": 1.0174,
+      "step": 10
+    },
+    {
+      "epoch": 0.026410564225690276,
+      "grad_norm": 2.4133969432396736,
+      "learning_rate": 2.5e-06,
+      "loss": 0.9751,
+      "step": 11
+    },
+    {
+      "epoch": 0.028811524609843937,
+      "grad_norm": 1.9294470459831854,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.9501,
+      "step": 12
+    },
+    {
+      "epoch": 0.031212484993997598,
+      "grad_norm": 1.6356435792199144,
+      "learning_rate": 3e-06,
+      "loss": 0.8964,
+      "step": 13
+    },
+    {
+      "epoch": 0.03361344537815126,
+      "grad_norm": 1.4259278773927042,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.9262,
+      "step": 14
+    },
+    {
+      "epoch": 0.03601440576230492,
+      "grad_norm": 1.4029510526993147,
+      "learning_rate": 3.5e-06,
+      "loss": 0.8773,
+      "step": 15
+    },
+    {
+      "epoch": 0.03841536614645858,
+      "grad_norm": 1.3072701950591272,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.8863,
+      "step": 16
+    },
+    {
+      "epoch": 0.04081632653061224,
+      "grad_norm": 1.3415661249879083,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.8391,
+      "step": 17
+    },
+    {
+      "epoch": 0.04321728691476591,
+      "grad_norm": 1.2551228414808617,
+      "learning_rate": 4.25e-06,
+      "loss": 0.8391,
+      "step": 18
+    },
+    {
+      "epoch": 0.04561824729891957,
+      "grad_norm": 1.4822117298636246,
+      "learning_rate": 4.5e-06,
+      "loss": 0.9249,
+      "step": 19
+    },
+    {
+      "epoch": 0.04801920768307323,
+      "grad_norm": 2.293387776769893,
+      "learning_rate": 4.75e-06,
+      "loss": 0.8683,
+      "step": 20
+    },
+    {
+      "epoch": 0.05042016806722689,
+      "grad_norm": 1.276527897418368,
+      "learning_rate": 5e-06,
+      "loss": 0.8863,
+      "step": 21
+    },
+    {
+      "epoch": 0.05282112845138055,
+      "grad_norm": 1.205563120747159,
+      "learning_rate": 5.2500000000000006e-06,
+      "loss": 0.8641,
+      "step": 22
+    },
+    {
+      "epoch": 0.055222088835534214,
+      "grad_norm": 1.2341891463491723,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.8741,
+      "step": 23
+    },
+    {
+      "epoch": 0.057623049219687875,
+      "grad_norm": 1.194537311448108,
+      "learning_rate": 5.75e-06,
+      "loss": 0.806,
+      "step": 24
+    },
+    {
+      "epoch": 0.060024009603841535,
+      "grad_norm": 1.2485956345196676,
+      "learning_rate": 6e-06,
+      "loss": 0.8281,
+      "step": 25
+    },
+    {
+      "epoch": 0.062424969987995196,
+      "grad_norm": 1.1779315877632301,
+      "learning_rate": 6.25e-06,
+      "loss": 0.8552,
+      "step": 26
+    },
+    {
+      "epoch": 0.06482593037214886,
+      "grad_norm": 1.4180731528009602,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.8506,
+      "step": 27
+    },
+    {
+      "epoch": 0.06722689075630252,
+      "grad_norm": 3.041894752350215,
+      "learning_rate": 6.750000000000001e-06,
+      "loss": 0.8885,
+      "step": 28
+    },
+    {
+      "epoch": 0.06962785114045618,
+      "grad_norm": 1.1860989270253326,
+      "learning_rate": 7e-06,
+      "loss": 0.8331,
+      "step": 29
+    },
+    {
+      "epoch": 0.07202881152460984,
+      "grad_norm": 1.206902242271321,
+      "learning_rate": 7.25e-06,
+      "loss": 0.8206,
+      "step": 30
+    },
+    {
+      "epoch": 0.0744297719087635,
+      "grad_norm": 1.1671538322004624,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.8167,
+      "step": 31
+    },
+    {
+      "epoch": 0.07683073229291716,
+      "grad_norm": 1.0422949222046782,
+      "learning_rate": 7.75e-06,
+      "loss": 0.8007,
+      "step": 32
+    },
+    {
+      "epoch": 0.07923169267707082,
+      "grad_norm": 1.8592446673491745,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.8507,
+      "step": 33
+    },
+    {
+      "epoch": 0.08163265306122448,
+      "grad_norm": 1.6612165554220881,
+      "learning_rate": 8.25e-06,
+      "loss": 0.8508,
+      "step": 34
+    },
+    {
+      "epoch": 0.08403361344537816,
+      "grad_norm": 1.0713315873933724,
+      "learning_rate": 8.5e-06,
+      "loss": 0.8605,
+      "step": 35
+    },
+    {
+      "epoch": 0.08643457382953182,
+      "grad_norm": 1.1490421409260716,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.8529,
+      "step": 36
+    },
+    {
+      "epoch": 0.08883553421368548,
+      "grad_norm": 1.1077552226131402,
+      "learning_rate": 9e-06,
+      "loss": 0.8251,
+      "step": 37
+    },
+    {
+      "epoch": 0.09123649459783914,
+      "grad_norm": 0.9873238748489626,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.7978,
+      "step": 38
+    },
+    {
+      "epoch": 0.0936374549819928,
+      "grad_norm": 1.3452579503150184,
+      "learning_rate": 9.5e-06,
+      "loss": 0.8637,
+      "step": 39
+    },
+    {
+      "epoch": 0.09603841536614646,
+      "grad_norm": 1.4631068406573169,
+      "learning_rate": 9.75e-06,
+      "loss": 0.8023,
+      "step": 40
+    },
+    {
+      "epoch": 0.09843937575030012,
+      "grad_norm": 1.265806360917147,
+      "learning_rate": 1e-05,
+      "loss": 0.8345,
+      "step": 41
+    },
+    {
+      "epoch": 0.10084033613445378,
+      "grad_norm": 1.1943330814308766,
+      "learning_rate": 9.999960664124435e-06,
+      "loss": 0.8704,
+      "step": 42
+    },
+    {
+      "epoch": 0.10324129651860744,
+      "grad_norm": 1.061823017754904,
+      "learning_rate": 9.999842657116667e-06,
+      "loss": 0.8463,
+      "step": 43
+    },
+    {
+      "epoch": 0.1056422569027611,
+      "grad_norm": 1.0296226499358687,
+      "learning_rate": 9.999645980833454e-06,
+      "loss": 0.825,
+      "step": 44
+    },
+    {
+      "epoch": 0.10804321728691477,
+      "grad_norm": 1.046275065503394,
+      "learning_rate": 9.999370638369377e-06,
+      "loss": 0.8341,
+      "step": 45
+    },
+    {
+      "epoch": 0.11044417767106843,
+      "grad_norm": 1.0958287099043753,
+      "learning_rate": 9.999016634056764e-06,
+      "loss": 0.7894,
+      "step": 46
+    },
+    {
+      "epoch": 0.11284513805522209,
+      "grad_norm": 1.030922973342192,
+      "learning_rate": 9.998583973465647e-06,
+      "loss": 0.787,
+      "step": 47
+    },
+    {
+      "epoch": 0.11524609843937575,
+      "grad_norm": 1.2421669534880109,
+      "learning_rate": 9.998072663403657e-06,
+      "loss": 0.7978,
+      "step": 48
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 1.3970247274710101,
+      "learning_rate": 9.997482711915926e-06,
+      "loss": 0.7801,
+      "step": 49
+    },
+    {
+      "epoch": 0.12004801920768307,
+      "grad_norm": 0.9723742502152596,
+      "learning_rate": 9.99681412828496e-06,
+      "loss": 0.7895,
+      "step": 50
+    },
+    {
+      "epoch": 0.12244897959183673,
+      "grad_norm": 1.2807089303778385,
+      "learning_rate": 9.996066923030484e-06,
+      "loss": 0.8543,
+      "step": 51
+    },
+    {
+      "epoch": 0.12484993997599039,
+      "grad_norm": 1.27637006864039,
+      "learning_rate": 9.99524110790929e-06,
+      "loss": 0.8535,
+      "step": 52
+    },
+    {
+      "epoch": 0.12725090036014405,
+      "grad_norm": 0.9583150465246295,
+      "learning_rate": 9.994336695915041e-06,
+      "loss": 0.7819,
+      "step": 53
+    },
+    {
+      "epoch": 0.12965186074429771,
+      "grad_norm": 1.066121996350333,
+      "learning_rate": 9.993353701278072e-06,
+      "loss": 0.7962,
+      "step": 54
+    },
+    {
+      "epoch": 0.13205282112845138,
+      "grad_norm": 1.0321017551908462,
+      "learning_rate": 9.992292139465166e-06,
+      "loss": 0.794,
+      "step": 55
+    },
+    {
+      "epoch": 0.13445378151260504,
+      "grad_norm": 1.1056060968463626,
+      "learning_rate": 9.991152027179307e-06,
+      "loss": 0.8431,
+      "step": 56
+    },
+    {
+      "epoch": 0.1368547418967587,
+      "grad_norm": 1.6782765780764348,
+      "learning_rate": 9.989933382359423e-06,
+      "loss": 0.8303,
+      "step": 57
+    },
+    {
+      "epoch": 0.13925570228091236,
+      "grad_norm": 1.1967331617073076,
+      "learning_rate": 9.988636224180097e-06,
+      "loss": 0.7792,
+      "step": 58
+    },
+    {
+      "epoch": 0.14165666266506602,
+      "grad_norm": 1.8049549266124272,
+      "learning_rate": 9.987260573051268e-06,
+      "loss": 0.7988,
+      "step": 59
+    },
+    {
+      "epoch": 0.14405762304921968,
+      "grad_norm": 1.046695584455822,
+      "learning_rate": 9.985806450617916e-06,
+      "loss": 0.8037,
+      "step": 60
+    },
+    {
+      "epoch": 0.14645858343337334,
+      "grad_norm": 1.2121335142230327,
+      "learning_rate": 9.984273879759713e-06,
+      "loss": 0.8515,
+      "step": 61
+    },
+    {
+      "epoch": 0.148859543817527,
+      "grad_norm": 1.3076719515024817,
+      "learning_rate": 9.982662884590662e-06,
+      "loss": 0.8178,
+      "step": 62
+    },
+    {
+      "epoch": 0.15126050420168066,
+      "grad_norm": 1.2564921607093686,
+      "learning_rate": 9.980973490458728e-06,
+      "loss": 0.8399,
+      "step": 63
+    },
+    {
+      "epoch": 0.15366146458583432,
+      "grad_norm": 1.0641383621124019,
+      "learning_rate": 9.97920572394543e-06,
+      "loss": 0.7677,
+      "step": 64
+    },
+    {
+      "epoch": 0.15606242496998798,
+      "grad_norm": 1.1281926663515902,
+      "learning_rate": 9.977359612865424e-06,
+      "loss": 0.7974,
+      "step": 65
+    },
+    {
+      "epoch": 0.15846338535414164,
+      "grad_norm": 1.0094739962757735,
+      "learning_rate": 9.975435186266069e-06,
+      "loss": 0.7676,
+      "step": 66
+    },
+    {
+      "epoch": 0.1608643457382953,
+      "grad_norm": 1.3602443348603426,
+      "learning_rate": 9.973432474426968e-06,
+      "loss": 0.8171,
+      "step": 67
+    },
+    {
+      "epoch": 0.16326530612244897,
+      "grad_norm": 0.9792328660506742,
+      "learning_rate": 9.971351508859488e-06,
+      "loss": 0.7561,
+      "step": 68
+    },
+    {
+      "epoch": 0.16566626650660263,
+      "grad_norm": 1.397240469672876,
+      "learning_rate": 9.969192322306271e-06,
+      "loss": 0.8365,
+      "step": 69
+    },
+    {
+      "epoch": 0.16806722689075632,
+      "grad_norm": 1.3993747618987225,
+      "learning_rate": 9.966954948740717e-06,
+      "loss": 0.7281,
+      "step": 70
+    },
+    {
+      "epoch": 0.17046818727490998,
+      "grad_norm": 1.0666295786391167,
+      "learning_rate": 9.964639423366442e-06,
+      "loss": 0.8563,
+      "step": 71
+    },
+    {
+      "epoch": 0.17286914765906364,
+      "grad_norm": 1.6098094871731998,
+      "learning_rate": 9.962245782616734e-06,
+      "loss": 0.8027,
+      "step": 72
+    },
+    {
+      "epoch": 0.1752701080432173,
+      "grad_norm": 1.0029300637530163,
+      "learning_rate": 9.959774064153977e-06,
+      "loss": 0.7768,
+      "step": 73
+    },
+    {
+      "epoch": 0.17767106842737096,
+      "grad_norm": 1.1557098402816255,
+      "learning_rate": 9.957224306869053e-06,
+      "loss": 0.7915,
+      "step": 74
+    },
+    {
+      "epoch": 0.18007202881152462,
+      "grad_norm": 1.1411436589531005,
+      "learning_rate": 9.954596550880735e-06,
+      "loss": 0.8207,
+      "step": 75
+    },
+    {
+      "epoch": 0.18247298919567828,
+      "grad_norm": 0.9898519262158086,
+      "learning_rate": 9.951890837535058e-06,
+      "loss": 0.8303,
+      "step": 76
+    },
+    {
+      "epoch": 0.18487394957983194,
+      "grad_norm": 1.015734040263864,
+      "learning_rate": 9.949107209404664e-06,
+      "loss": 0.7928,
+      "step": 77
+    },
+    {
+      "epoch": 0.1872749099639856,
+      "grad_norm": 1.0314783989677847,
+      "learning_rate": 9.946245710288132e-06,
+      "loss": 0.7407,
+      "step": 78
+    },
+    {
+      "epoch": 0.18967587034813926,
+      "grad_norm": 2.1401440917485943,
+      "learning_rate": 9.94330638520929e-06,
+      "loss": 0.7775,
+      "step": 79
+    },
+    {
+      "epoch": 0.19207683073229292,
+      "grad_norm": 0.9890405160925984,
+      "learning_rate": 9.940289280416509e-06,
+      "loss": 0.7566,
+      "step": 80
+    },
+    {
+      "epoch": 0.19447779111644659,
+      "grad_norm": 0.9739543182418596,
+      "learning_rate": 9.937194443381972e-06,
+      "loss": 0.8159,
+      "step": 81
+    },
+    {
+      "epoch": 0.19687875150060025,
+      "grad_norm": 1.1138048276749253,
+      "learning_rate": 9.934021922800931e-06,
+      "loss": 0.7939,
+      "step": 82
+    },
+    {
+      "epoch": 0.1992797118847539,
+      "grad_norm": 1.0939436400008165,
+      "learning_rate": 9.930771768590934e-06,
+      "loss": 0.8144,
+      "step": 83
+    },
+    {
+      "epoch": 0.20168067226890757,
+      "grad_norm": 1.1293648489823769,
+      "learning_rate": 9.927444031891044e-06,
+      "loss": 0.7755,
+      "step": 84
+    },
+    {
+      "epoch": 0.20408163265306123,
+      "grad_norm": 1.0430331448777437,
+      "learning_rate": 9.924038765061042e-06,
+      "loss": 0.8215,
+      "step": 85
+    },
+    {
+      "epoch": 0.2064825930372149,
+      "grad_norm": 1.9537322284504008,
+      "learning_rate": 9.92055602168058e-06,
+      "loss": 0.7563,
+      "step": 86
+    },
+    {
+      "epoch": 0.20888355342136855,
+      "grad_norm": 1.3079086606300951,
+      "learning_rate": 9.916995856548371e-06,
+      "loss": 0.8157,
+      "step": 87
+    },
+    {
+      "epoch": 0.2112845138055222,
+      "grad_norm": 0.9221688539776796,
+      "learning_rate": 9.913358325681292e-06,
+      "loss": 0.782,
+      "step": 88
+    },
+    {
+      "epoch": 0.21368547418967587,
+      "grad_norm": 1.1087405466326585,
+      "learning_rate": 9.909643486313533e-06,
+      "loss": 0.8061,
+      "step": 89
+    },
+    {
+      "epoch": 0.21608643457382953,
+      "grad_norm": 1.2632508204228434,
+      "learning_rate": 9.905851396895679e-06,
+      "loss": 0.7788,
+      "step": 90
+    },
+    {
+      "epoch": 0.2184873949579832,
+      "grad_norm": 1.2841221080352947,
+      "learning_rate": 9.901982117093786e-06,
+      "loss": 0.7707,
+      "step": 91
+    },
+    {
+      "epoch": 0.22088835534213686,
+      "grad_norm": 0.9416571120286465,
+      "learning_rate": 9.898035707788462e-06,
+      "loss": 0.7386,
+      "step": 92
+    },
+    {
+      "epoch": 0.22328931572629052,
+      "grad_norm": 1.0394686320396263,
+      "learning_rate": 9.894012231073895e-06,
+      "loss": 0.7729,
+      "step": 93
+    },
+    {
+      "epoch": 0.22569027611044418,
+      "grad_norm": 1.1642999740100775,
+      "learning_rate": 9.889911750256873e-06,
+      "loss": 0.7667,
+      "step": 94
+    },
+    {
+      "epoch": 0.22809123649459784,
+      "grad_norm": 1.0191501234681113,
+      "learning_rate": 9.885734329855798e-06,
+      "loss": 0.7415,
+      "step": 95
+    },
+    {
+      "epoch": 0.2304921968787515,
+      "grad_norm": 1.1396215768269955,
+      "learning_rate": 9.881480035599667e-06,
+      "loss": 0.846,
+      "step": 96
+    },
+    {
+      "epoch": 0.23289315726290516,
+      "grad_norm": 1.4770430873772629,
+      "learning_rate": 9.877148934427037e-06,
+      "loss": 0.8105,
+      "step": 97
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 1.1245875213463206,
+      "learning_rate": 9.872741094484965e-06,
+      "loss": 0.7857,
+      "step": 98
+    },
+    {
+      "epoch": 0.23769507803121248,
+      "grad_norm": 0.9225248681792048,
+      "learning_rate": 9.868256585127956e-06,
+      "loss": 0.7292,
+      "step": 99
+    },
+    {
+      "epoch": 0.24009603841536614,
+      "grad_norm": 1.0574440145567323,
+      "learning_rate": 9.863695476916846e-06,
+      "loss": 0.8077,
+      "step": 100
+    },
+    {
+      "epoch": 0.2424969987995198,
+      "grad_norm": 1.1365137152954194,
+      "learning_rate": 9.859057841617709e-06,
+      "loss": 0.8066,
+      "step": 101
+    },
+    {
+      "epoch": 0.24489795918367346,
+      "grad_norm": 1.1942946403505255,
+      "learning_rate": 9.854343752200725e-06,
+      "loss": 0.8139,
+      "step": 102
+    },
+    {
+      "epoch": 0.24729891956782712,
+      "grad_norm": 1.0543479220950105,
+      "learning_rate": 9.849553282839025e-06,
+      "loss": 0.7701,
+      "step": 103
+    },
+    {
+      "epoch": 0.24969987995198079,
+      "grad_norm": 0.926743129553519,
+      "learning_rate": 9.844686508907538e-06,
+      "loss": 0.7927,
+      "step": 104
+    },
+    {
+      "epoch": 0.25210084033613445,
+      "grad_norm": 1.011488885523952,
+      "learning_rate": 9.839743506981783e-06,
+      "loss": 0.7914,
+      "step": 105
+    },
+    {
+      "epoch": 0.2545018007202881,
+      "grad_norm": 1.0064857869367891,
+      "learning_rate": 9.834724354836684e-06,
+      "loss": 0.7701,
+      "step": 106
+    },
+    {
+      "epoch": 0.25690276110444177,
+      "grad_norm": 1.2884610797254652,
+      "learning_rate": 9.829629131445342e-06,
+      "loss": 0.7797,
+      "step": 107
+    },
+    {
+      "epoch": 0.25930372148859543,
+      "grad_norm": 1.022923038777103,
+      "learning_rate": 9.824457916977785e-06,
+      "loss": 0.7563,
+      "step": 108
+    },
+    {
+      "epoch": 0.2617046818727491,
+      "grad_norm": 1.3622638931252318,
+      "learning_rate": 9.819210792799711e-06,
+      "loss": 0.791,
+      "step": 109
+    },
+    {
+      "epoch": 0.26410564225690275,
+      "grad_norm": 1.6261815582055081,
+      "learning_rate": 9.81388784147121e-06,
+      "loss": 0.7791,
+      "step": 110
+    },
+    {
+      "epoch": 0.2665066026410564,
+      "grad_norm": 1.3875850558476444,
+      "learning_rate": 9.808489146745466e-06,
+      "loss": 0.8065,
+      "step": 111
+    },
+    {
+      "epoch": 0.2689075630252101,
+      "grad_norm": 1.0566640937075868,
+      "learning_rate": 9.803014793567429e-06,
+      "loss": 0.7662,
+      "step": 112
+    },
+    {
+      "epoch": 0.27130852340936373,
+      "grad_norm": 0.9297104835810871,
+      "learning_rate": 9.797464868072489e-06,
+      "loss": 0.7338,
+      "step": 113
+    },
+    {
+      "epoch": 0.2737094837935174,
+      "grad_norm": 0.9992084910627275,
+      "learning_rate": 9.791839457585118e-06,
+      "loss": 0.7773,
+      "step": 114
+    },
+    {
+      "epoch": 0.27611044417767105,
+      "grad_norm": 1.1581522702903588,
+      "learning_rate": 9.786138650617494e-06,
+      "loss": 0.7929,
+      "step": 115
+    },
+    {
+      "epoch": 0.2785114045618247,
+      "grad_norm": 1.259081791921377,
+      "learning_rate": 9.780362536868113e-06,
+      "loss": 0.7279,
+      "step": 116
+    },
+    {
+      "epoch": 0.2809123649459784,
+      "grad_norm": 1.1510251918441672,
+      "learning_rate": 9.774511207220369e-06,
+      "loss": 0.7412,
+      "step": 117
+    },
+    {
+      "epoch": 0.28331332533013204,
+      "grad_norm": 0.9943579023159135,
+      "learning_rate": 9.768584753741134e-06,
+      "loss": 0.7927,
+      "step": 118
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 1.061487526635317,
+      "learning_rate": 9.762583269679304e-06,
+      "loss": 0.8363,
+      "step": 119
+    },
+    {
+      "epoch": 0.28811524609843936,
+      "grad_norm": 1.1215382733800197,
+      "learning_rate": 9.756506849464327e-06,
+      "loss": 0.8032,
+      "step": 120
+    },
+    {
+      "epoch": 0.290516206482593,
+      "grad_norm": 1.0985810622772545,
+      "learning_rate": 9.750355588704728e-06,
+      "loss": 0.8062,
+      "step": 121
+    },
+    {
+      "epoch": 0.2929171668667467,
+      "grad_norm": 1.0397097160893556,
+      "learning_rate": 9.744129584186599e-06,
+      "loss": 0.7867,
+      "step": 122
+    },
+    {
+      "epoch": 0.29531812725090034,
+      "grad_norm": 1.0985983029789874,
+      "learning_rate": 9.737828933872076e-06,
+      "loss": 0.7645,
+      "step": 123
+    },
+    {
+      "epoch": 0.297719087635054,
+      "grad_norm": 1.0341542778256945,
+      "learning_rate": 9.731453736897796e-06,
+      "loss": 0.7682,
+      "step": 124
+    },
+    {
+      "epoch": 0.30012004801920766,
+      "grad_norm": 0.9637556163656881,
+      "learning_rate": 9.725004093573343e-06,
+      "loss": 0.7819,
+      "step": 125
+    },
+    {
+      "epoch": 0.3025210084033613,
+      "grad_norm": 0.9476489500974398,
+      "learning_rate": 9.718480105379663e-06,
+      "loss": 0.7875,
+      "step": 126
+    },
+    {
+      "epoch": 0.304921968787515,
+      "grad_norm": 1.3085520795980778,
+      "learning_rate": 9.711881874967471e-06,
+      "loss": 0.7908,
+      "step": 127
+    },
+    {
+      "epoch": 0.30732292917166865,
+      "grad_norm": 0.9414082676216773,
+      "learning_rate": 9.705209506155635e-06,
+      "loss": 0.7454,
+      "step": 128
+    },
+    {
+      "epoch": 0.3097238895558223,
+      "grad_norm": 1.2351287282053665,
+      "learning_rate": 9.698463103929542e-06,
+      "loss": 0.7813,
+      "step": 129
+    },
+    {
+      "epoch": 0.31212484993997597,
+      "grad_norm": 1.021583176125914,
+      "learning_rate": 9.69164277443945e-06,
+      "loss": 0.8417,
+      "step": 130
+    },
+    {
+      "epoch": 0.31452581032412963,
+      "grad_norm": 1.4089615187590279,
+      "learning_rate": 9.68474862499881e-06,
+      "loss": 0.7358,
+      "step": 131
+    },
+    {
+      "epoch": 0.3169267707082833,
+      "grad_norm": 0.9398428308548782,
+      "learning_rate": 9.677780764082583e-06,
+      "loss": 0.7637,
+      "step": 132
+    },
+    {
+      "epoch": 0.31932773109243695,
+      "grad_norm": 1.1136331841332863,
+      "learning_rate": 9.670739301325534e-06,
+      "loss": 0.7723,
+      "step": 133
+    },
+    {
+      "epoch": 0.3217286914765906,
+      "grad_norm": 1.1992806853517075,
+      "learning_rate": 9.663624347520506e-06,
+      "loss": 0.7788,
+      "step": 134
+    },
+    {
+      "epoch": 0.3241296518607443,
+      "grad_norm": 1.311509502714469,
+      "learning_rate": 9.65643601461667e-06,
+      "loss": 0.7696,
+      "step": 135
+    },
+    {
+      "epoch": 0.32653061224489793,
+      "grad_norm": 0.9623376564123519,
+      "learning_rate": 9.649174415717776e-06,
+      "loss": 0.7414,
+      "step": 136
+    },
+    {
+      "epoch": 0.3289315726290516,
+      "grad_norm": 1.0486840444117425,
+      "learning_rate": 9.641839665080363e-06,
+      "loss": 0.7909,
+      "step": 137
+    },
+    {
+      "epoch": 0.33133253301320525,
+      "grad_norm": 1.172648282688611,
+      "learning_rate": 9.634431878111969e-06,
+      "loss": 0.7684,
+      "step": 138
+    },
+    {
+      "epoch": 0.33373349339735897,
+      "grad_norm": 1.3883931408734287,
+      "learning_rate": 9.626951171369306e-06,
+      "loss": 0.7824,
+      "step": 139
+    },
+    {
+      "epoch": 0.33613445378151263,
+      "grad_norm": 1.0459492696464887,
+      "learning_rate": 9.619397662556434e-06,
+      "loss": 0.7729,
+      "step": 140
+    },
+    {
+      "epoch": 0.3385354141656663,
+      "grad_norm": 1.090681276032822,
+      "learning_rate": 9.611771470522908e-06,
+      "loss": 0.78,
+      "step": 141
+    },
+    {
+      "epoch": 0.34093637454981995,
+      "grad_norm": 1.077439313381409,
+      "learning_rate": 9.604072715261902e-06,
+      "loss": 0.7746,
+      "step": 142
+    },
+    {
+      "epoch": 0.3433373349339736,
+      "grad_norm": 1.0119948107455536,
+      "learning_rate": 9.596301517908329e-06,
+      "loss": 0.7725,
+      "step": 143
+    },
+    {
+      "epoch": 0.3457382953181273,
+      "grad_norm": 1.0261650566322091,
+      "learning_rate": 9.588458000736929e-06,
+      "loss": 0.8126,
+      "step": 144
+    },
+    {
+      "epoch": 0.34813925570228094,
+      "grad_norm": 3.2915968630848296,
+      "learning_rate": 9.580542287160348e-06,
+      "loss": 0.8087,
+      "step": 145
+    },
+    {
+      "epoch": 0.3505402160864346,
+      "grad_norm": 1.1260052143600965,
+      "learning_rate": 9.572554501727198e-06,
+      "loss": 0.7626,
+      "step": 146
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 1.1992270704321544,
+      "learning_rate": 9.564494770120089e-06,
+      "loss": 0.7942,
+      "step": 147
+    },
+    {
+      "epoch": 0.3553421368547419,
+      "grad_norm": 0.8934279508188496,
+      "learning_rate": 9.556363219153664e-06,
+      "loss": 0.7121,
+      "step": 148
+    },
+    {
+      "epoch": 0.3577430972388956,
+      "grad_norm": 1.627029857609285,
+      "learning_rate": 9.548159976772593e-06,
+      "loss": 0.8305,
+      "step": 149
+    },
+    {
+      "epoch": 0.36014405762304924,
+      "grad_norm": 1.2904112895903999,
+      "learning_rate": 9.539885172049563e-06,
+      "loss": 0.7903,
+      "step": 150
+    },
+    {
+      "epoch": 0.3625450180072029,
+      "grad_norm": 0.9013735933905231,
+      "learning_rate": 9.531538935183252e-06,
+      "loss": 0.7532,
+      "step": 151
+    },
+    {
+      "epoch": 0.36494597839135656,
+      "grad_norm": 1.1957280125341616,
+      "learning_rate": 9.52312139749627e-06,
+      "loss": 0.7654,
+      "step": 152
+    },
+    {
+      "epoch": 0.3673469387755102,
+      "grad_norm": 1.1864890770385386,
+      "learning_rate": 9.514632691433108e-06,
+      "loss": 0.7774,
+      "step": 153
+    },
+    {
+      "epoch": 0.3697478991596639,
+      "grad_norm": 1.2602126687460933,
+      "learning_rate": 9.506072950558036e-06,
+      "loss": 0.7308,
+      "step": 154
+    },
+    {
+      "epoch": 0.37214885954381755,
+      "grad_norm": 0.9838354835983466,
+      "learning_rate": 9.497442309553017e-06,
+      "loss": 0.8117,
+      "step": 155
+    },
+    {
+      "epoch": 0.3745498199279712,
+      "grad_norm": 1.0381143846448244,
+      "learning_rate": 9.488740904215578e-06,
+      "loss": 0.7949,
+      "step": 156
+    },
+    {
+      "epoch": 0.37695078031212487,
+      "grad_norm": 1.0974151220229578,
+      "learning_rate": 9.47996887145668e-06,
+      "loss": 0.8096,
+      "step": 157
+    },
+    {
+      "epoch": 0.3793517406962785,
+      "grad_norm": 1.0644973845655654,
+      "learning_rate": 9.471126349298557e-06,
+      "loss": 0.769,
+      "step": 158
+    },
+    {
+      "epoch": 0.3817527010804322,
+      "grad_norm": 1.0805369220592416,
+      "learning_rate": 9.46221347687255e-06,
+      "loss": 0.7528,
+      "step": 159
+    },
+    {
+      "epoch": 0.38415366146458585,
+      "grad_norm": 1.0244991912422095,
+      "learning_rate": 9.453230394416914e-06,
+      "loss": 0.7883,
+      "step": 160
+    },
+    {
+      "epoch": 0.3865546218487395,
+      "grad_norm": 0.9661371626174572,
+      "learning_rate": 9.444177243274619e-06,
+      "loss": 0.784,
+      "step": 161
+    },
+    {
+      "epoch": 0.38895558223289317,
+      "grad_norm": 1.3148287207686615,
+      "learning_rate": 9.43505416589111e-06,
+      "loss": 0.7656,
+      "step": 162
+    },
+    {
+      "epoch": 0.39135654261704683,
+      "grad_norm": 1.1619055283617015,
+      "learning_rate": 9.425861305812083e-06,
+      "loss": 0.7824,
+      "step": 163
+    },
+    {
+      "epoch": 0.3937575030012005,
+      "grad_norm": 4.651513081191553,
+      "learning_rate": 9.416598807681221e-06,
+      "loss": 0.8066,
+      "step": 164
+    },
+    {
+      "epoch": 0.39615846338535415,
+      "grad_norm": 0.9479957292008633,
+      "learning_rate": 9.40726681723791e-06,
+      "loss": 0.7651,
+      "step": 165
+    },
+    {
+      "epoch": 0.3985594237695078,
+      "grad_norm": 1.1703185690222349,
+      "learning_rate": 9.397865481314959e-06,
+      "loss": 0.7656,
+      "step": 166
+    },
+    {
+      "epoch": 0.4009603841536615,
+      "grad_norm": 1.0368151311067004,
+      "learning_rate": 9.388394947836278e-06,
+      "loss": 0.7917,
+      "step": 167
+    },
+    {
+      "epoch": 0.40336134453781514,
+      "grad_norm": 2.298562774870953,
+      "learning_rate": 9.37885536581456e-06,
+      "loss": 0.762,
+      "step": 168
+    },
+    {
+      "epoch": 0.4057623049219688,
+      "grad_norm": 1.1062217919442925,
+      "learning_rate": 9.369246885348926e-06,
+      "loss": 0.7585,
+      "step": 169
+    },
+    {
+      "epoch": 0.40816326530612246,
+      "grad_norm": 1.173156390813293,
+      "learning_rate": 9.359569657622573e-06,
+      "loss": 0.7797,
+      "step": 170
+    },
+    {
+      "epoch": 0.4105642256902761,
+      "grad_norm": 1.109809941134885,
+      "learning_rate": 9.349823834900396e-06,
+      "loss": 0.7588,
+      "step": 171
+    },
+    {
+      "epoch": 0.4129651860744298,
+      "grad_norm": 1.2813177351957437,
+      "learning_rate": 9.340009570526578e-06,
+      "loss": 0.7616,
+      "step": 172
+    },
+    {
+      "epoch": 0.41536614645858344,
+      "grad_norm": 1.0591641661573232,
+      "learning_rate": 9.330127018922195e-06,
+      "loss": 0.7564,
+      "step": 173
+    },
+    {
+      "epoch": 0.4177671068427371,
+      "grad_norm": 2.376914868004014,
+      "learning_rate": 9.320176335582772e-06,
+      "loss": 0.7366,
+      "step": 174
+    },
+    {
+      "epoch": 0.42016806722689076,
+      "grad_norm": 0.9902616640042359,
+      "learning_rate": 9.310157677075847e-06,
+      "loss": 0.7429,
+      "step": 175
+    },
+    {
+      "epoch": 0.4225690276110444,
+      "grad_norm": 1.6847159971307049,
+      "learning_rate": 9.300071201038503e-06,
+      "loss": 0.7825,
+      "step": 176
+    },
+    {
+      "epoch": 0.4249699879951981,
+      "grad_norm": 0.9942879778667437,
+      "learning_rate": 9.289917066174887e-06,
+      "loss": 0.7318,
+      "step": 177
+    },
+    {
+      "epoch": 0.42737094837935174,
+      "grad_norm": 1.1742665576935563,
+      "learning_rate": 9.27969543225371e-06,
+      "loss": 0.7623,
+      "step": 178
+    },
+    {
+      "epoch": 0.4297719087635054,
+      "grad_norm": 1.2676535006378862,
+      "learning_rate": 9.269406460105742e-06,
+      "loss": 0.7583,
+      "step": 179
+    },
+    {
+      "epoch": 0.43217286914765907,
+      "grad_norm": 1.678798458940733,
+      "learning_rate": 9.259050311621274e-06,
+      "loss": 0.7881,
+      "step": 180
+    },
+    {
+      "epoch": 0.4345738295318127,
+      "grad_norm": 1.1441951682044669,
+      "learning_rate": 9.248627149747573e-06,
+      "loss": 0.7455,
+      "step": 181
+    },
+    {
+      "epoch": 0.4369747899159664,
+      "grad_norm": 1.0055990609873005,
+      "learning_rate": 9.238137138486318e-06,
+      "loss": 0.7613,
+      "step": 182
+    },
+    {
+      "epoch": 0.43937575030012005,
+      "grad_norm": 1.0631388873420178,
+      "learning_rate": 9.227580442891022e-06,
+      "loss": 0.7528,
+      "step": 183
+    },
+    {
+      "epoch": 0.4417767106842737,
+      "grad_norm": 1.0811446737171562,
+      "learning_rate": 9.21695722906443e-06,
+      "loss": 0.7988,
+      "step": 184
+    },
+    {
+      "epoch": 0.44417767106842737,
+      "grad_norm": 3.77472894386223,
+      "learning_rate": 9.206267664155906e-06,
+      "loss": 0.7669,
+      "step": 185
+    },
+    {
+      "epoch": 0.44657863145258103,
+      "grad_norm": 2.5440967403332833,
+      "learning_rate": 9.195511916358813e-06,
+      "loss": 0.7478,
+      "step": 186
+    },
+    {
+      "epoch": 0.4489795918367347,
+      "grad_norm": 1.1995260606009623,
+      "learning_rate": 9.18469015490785e-06,
+      "loss": 0.7637,
+      "step": 187
+    },
+    {
+      "epoch": 0.45138055222088835,
+      "grad_norm": 1.087601696720053,
+      "learning_rate": 9.173802550076402e-06,
+      "loss": 0.7122,
+      "step": 188
+    },
+    {
+      "epoch": 0.453781512605042,
+      "grad_norm": 0.859793070085179,
+      "learning_rate": 9.162849273173857e-06,
+      "loss": 0.7237,
+      "step": 189
+    },
+    {
+      "epoch": 0.4561824729891957,
+      "grad_norm": 0.909405499890903,
+      "learning_rate": 9.151830496542912e-06,
+      "loss": 0.7413,
+      "step": 190
+    },
+    {
+      "epoch": 0.45858343337334934,
+      "grad_norm": 1.1698481579798348,
+      "learning_rate": 9.140746393556853e-06,
+      "loss": 0.7686,
+      "step": 191
+    },
+    {
+      "epoch": 0.460984393757503,
+      "grad_norm": 1.4005910755153028,
+      "learning_rate": 9.129597138616845e-06,
+      "loss": 0.7301,
+      "step": 192
+    },
+    {
+      "epoch": 0.46338535414165666,
+      "grad_norm": 6.016086016376545,
+      "learning_rate": 9.118382907149164e-06,
+      "loss": 0.7871,
+      "step": 193
+    },
+    {
+      "epoch": 0.4657863145258103,
+      "grad_norm": 0.9147643453009918,
+      "learning_rate": 9.107103875602458e-06,
+      "loss": 0.7753,
+      "step": 194
+    },
+    {
+      "epoch": 0.468187274909964,
+      "grad_norm": 1.1762754554445651,
+      "learning_rate": 9.09576022144496e-06,
+      "loss": 0.7764,
+      "step": 195
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 1.175295130468933,
+      "learning_rate": 9.084352123161695e-06,
+      "loss": 0.7912,
+      "step": 196
+    },
+    {
+      "epoch": 0.4729891956782713,
+      "grad_norm": 1.0739318228013277,
+      "learning_rate": 9.07287976025168e-06,
+      "loss": 0.762,
+      "step": 197
+    },
+    {
+      "epoch": 0.47539015606242496,
+      "grad_norm": 0.9934440139468771,
+      "learning_rate": 9.061343313225088e-06,
+      "loss": 0.8003,
+      "step": 198
+    },
+    {
+      "epoch": 0.4777911164465786,
+      "grad_norm": 1.0768642208250674,
+      "learning_rate": 9.04974296360042e-06,
+      "loss": 0.7659,
+      "step": 199
+    },
+    {
+      "epoch": 0.4801920768307323,
+      "grad_norm": 1.1273977823886203,
+      "learning_rate": 9.038078893901634e-06,
+      "loss": 0.7992,
+      "step": 200
+    },
+    {
+      "epoch": 0.48259303721488594,
+      "grad_norm": 1.1987531653944965,
+      "learning_rate": 9.026351287655294e-06,
+      "loss": 0.7932,
+      "step": 201
+    },
+    {
+      "epoch": 0.4849939975990396,
+      "grad_norm": 1.4257297426073774,
+      "learning_rate": 9.014560329387661e-06,
+      "loss": 0.7372,
+      "step": 202
+    },
+    {
+      "epoch": 0.48739495798319327,
+      "grad_norm": 0.8935161112029553,
+      "learning_rate": 9.002706204621802e-06,
+      "loss": 0.7701,
+      "step": 203
+    },
+    {
+      "epoch": 0.4897959183673469,
+      "grad_norm": 1.1445504107150832,
+      "learning_rate": 8.99078909987467e-06,
+      "loss": 0.7937,
+      "step": 204
+    },
+    {
+      "epoch": 0.4921968787515006,
+      "grad_norm": 0.9975014914511301,
+      "learning_rate": 8.978809202654161e-06,
+      "loss": 0.7045,
+      "step": 205
+    },
+    {
+      "epoch": 0.49459783913565425,
+      "grad_norm": 1.566224538850411,
+      "learning_rate": 8.966766701456177e-06,
+      "loss": 0.7665,
+      "step": 206
+    },
+    {
+      "epoch": 0.4969987995198079,
+      "grad_norm": 1.0336200728044336,
+      "learning_rate": 8.954661785761648e-06,
+      "loss": 0.7635,
+      "step": 207
+    },
+    {
+      "epoch": 0.49939975990396157,
+      "grad_norm": 1.2720894224504704,
+      "learning_rate": 8.942494646033555e-06,
+      "loss": 0.7546,
+      "step": 208
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 832,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 208,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3314997980286157e+17,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-208/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:792152ea19e5a2c7444f7ec09222de7bca6342b94e2b0276449c3beea8acf729
+size 9105

checkpoint-208/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``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``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)