Upload folder using huggingface_hub

40_128_e5_3e-5/checkpoint-1035/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+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

40_128_e5_3e-5/checkpoint-1035/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

40_128_e5_3e-5/checkpoint-1035/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a51fafaefc824376d0517620d0a8c8a713d0358d21a7fa70450740ea3607c054
+size 791751704

40_128_e5_3e-5/checkpoint-1035/latest

@@ -0,0 +1 @@
+global_step1035

40_128_e5_3e-5/checkpoint-1035/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ebc178398316a8227ff7c3d9121aaef795982d56d45bb5af7bccce28d5849de
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5cf321e3a5cf312dfca051cff530d1143f3f44768fd32190841a7d8b0a2051f
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:672dac46335c4722c4422e8a52cafff669105e544046ef68e5b20fd8157bd44b
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4c86662e98c8d7f1cd8eb25b4d464db6f05055392652f17604955593ac91a51
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f00033a411abe702e0671f7da918a2bcbb824f65aa54c538902d5f8b7c0e96c
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9c84fa0dd2ef27649e5769753864d6aba11c5d3053c0b708057fd828c2d73e7
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb07fc431f0962f9d20be17bb91392612be5a79274d9bd6f45575b35293a48da
+size 15920

40_128_e5_3e-5/checkpoint-1035/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a007c9faeee33467e5449ee47e16853af8a33b81ad3d625a008b13ea444aa0e
+size 15920

40_128_e5_3e-5/checkpoint-1035/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:226f327be7cf3971e1ca108cf49105cb0041e9467b2bba962df5efce9f750632
+size 1064

40_128_e5_3e-5/checkpoint-1035/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

40_128_e5_3e-5/checkpoint-1035/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

40_128_e5_3e-5/checkpoint-1035/trainer_state.json

@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1035,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01451378809869376,
+      "grad_norm": 1.1063576936721802,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 1.3635,
+      "step": 5
+    },
+    {
+      "epoch": 0.02902757619738752,
+      "grad_norm": 0.8837594389915466,
+      "learning_rate": 3.103448275862069e-06,
+      "loss": 1.3281,
+      "step": 10
+    },
+    {
+      "epoch": 0.04354136429608128,
+      "grad_norm": 0.6033729314804077,
+      "learning_rate": 4.827586206896552e-06,
+      "loss": 1.37,
+      "step": 15
+    },
+    {
+      "epoch": 0.05805515239477504,
+      "grad_norm": 0.6915555000305176,
+      "learning_rate": 6.551724137931035e-06,
+      "loss": 1.2765,
+      "step": 20
+    },
+    {
+      "epoch": 0.07256894049346879,
+      "grad_norm": 0.7305431365966797,
+      "learning_rate": 8.275862068965517e-06,
+      "loss": 1.2649,
+      "step": 25
+    },
+    {
+      "epoch": 0.08708272859216255,
+      "grad_norm": 0.47839707136154175,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 1.2497,
+      "step": 30
+    },
+    {
+      "epoch": 0.10159651669085631,
+      "grad_norm": 0.5838127136230469,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 1.2576,
+      "step": 35
+    },
+    {
+      "epoch": 0.11611030478955008,
+      "grad_norm": 0.45399269461631775,
+      "learning_rate": 1.3448275862068966e-05,
+      "loss": 1.2541,
+      "step": 40
+    },
+    {
+      "epoch": 0.13062409288824384,
+      "grad_norm": 0.45961660146713257,
+      "learning_rate": 1.517241379310345e-05,
+      "loss": 1.1878,
+      "step": 45
+    },
+    {
+      "epoch": 0.14513788098693758,
+      "grad_norm": 0.5068319439888,
+      "learning_rate": 1.6896551724137932e-05,
+      "loss": 1.2369,
+      "step": 50
+    },
+    {
+      "epoch": 0.15965166908563136,
+      "grad_norm": 0.44543999433517456,
+      "learning_rate": 1.8620689655172415e-05,
+      "loss": 1.2221,
+      "step": 55
+    },
+    {
+      "epoch": 0.1741654571843251,
+      "grad_norm": 0.47043296694755554,
+      "learning_rate": 2.0344827586206894e-05,
+      "loss": 1.2211,
+      "step": 60
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.5043661594390869,
+      "learning_rate": 2.206896551724138e-05,
+      "loss": 1.1938,
+      "step": 65
+    },
+    {
+      "epoch": 0.20319303338171263,
+      "grad_norm": 0.4813358783721924,
+      "learning_rate": 2.3793103448275862e-05,
+      "loss": 1.2066,
+      "step": 70
+    },
+    {
+      "epoch": 0.21770682148040638,
+      "grad_norm": 0.44516244530677795,
+      "learning_rate": 2.5517241379310345e-05,
+      "loss": 1.175,
+      "step": 75
+    },
+    {
+      "epoch": 0.23222060957910015,
+      "grad_norm": 0.5195889472961426,
+      "learning_rate": 2.7241379310344827e-05,
+      "loss": 1.1696,
+      "step": 80
+    },
+    {
+      "epoch": 0.2467343976777939,
+      "grad_norm": 0.45658233761787415,
+      "learning_rate": 2.8965517241379313e-05,
+      "loss": 1.1516,
+      "step": 85
+    },
+    {
+      "epoch": 0.2612481857764877,
+      "grad_norm": 0.4927058517932892,
+      "learning_rate": 2.999988964483289e-05,
+      "loss": 1.1275,
+      "step": 90
+    },
+    {
+      "epoch": 0.2757619738751814,
+      "grad_norm": 0.5645100474357605,
+      "learning_rate": 2.9998648167850816e-05,
+      "loss": 1.1119,
+      "step": 95
+    },
+    {
+      "epoch": 0.29027576197387517,
+      "grad_norm": 0.5247628688812256,
+      "learning_rate": 2.9996027384477114e-05,
+      "loss": 1.1642,
+      "step": 100
+    },
+    {
+      "epoch": 0.3047895500725689,
+      "grad_norm": 0.562659740447998,
+      "learning_rate": 2.999202753572439e-05,
+      "loss": 1.1142,
+      "step": 105
+    },
+    {
+      "epoch": 0.3193033381712627,
+      "grad_norm": 0.5908432602882385,
+      "learning_rate": 2.998664898942692e-05,
+      "loss": 1.0488,
+      "step": 110
+    },
+    {
+      "epoch": 0.33381712626995647,
+      "grad_norm": 0.6552732586860657,
+      "learning_rate": 2.997989224020684e-05,
+      "loss": 1.1,
+      "step": 115
+    },
+    {
+      "epoch": 0.3483309143686502,
+      "grad_norm": 0.5694405436515808,
+      "learning_rate": 2.9971757909428642e-05,
+      "loss": 1.0119,
+      "step": 120
+    },
+    {
+      "epoch": 0.36284470246734396,
+      "grad_norm": 0.6327081918716431,
+      "learning_rate": 2.996224674514203e-05,
+      "loss": 1.0156,
+      "step": 125
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.60078364610672,
+      "learning_rate": 2.995135962201315e-05,
+      "loss": 1.0178,
+      "step": 130
+    },
+    {
+      "epoch": 0.3918722786647315,
+      "grad_norm": 0.7008715867996216,
+      "learning_rate": 2.993909754124413e-05,
+      "loss": 0.9871,
+      "step": 135
+    },
+    {
+      "epoch": 0.40638606676342526,
+      "grad_norm": 0.6301591396331787,
+      "learning_rate": 2.9925461630481022e-05,
+      "loss": 1.0024,
+      "step": 140
+    },
+    {
+      "epoch": 0.420899854862119,
+      "grad_norm": 0.6791720390319824,
+      "learning_rate": 2.9910453143710096e-05,
+      "loss": 0.9605,
+      "step": 145
+    },
+    {
+      "epoch": 0.43541364296081275,
+      "grad_norm": 0.650157630443573,
+      "learning_rate": 2.9894073461142523e-05,
+      "loss": 1.0019,
+      "step": 150
+    },
+    {
+      "epoch": 0.44992743105950656,
+      "grad_norm": 0.7557404637336731,
+      "learning_rate": 2.9876324089087442e-05,
+      "loss": 0.9906,
+      "step": 155
+    },
+    {
+      "epoch": 0.4644412191582003,
+      "grad_norm": 0.74793940782547,
+      "learning_rate": 2.9857206659813447e-05,
+      "loss": 1.0025,
+      "step": 160
+    },
+    {
+      "epoch": 0.47895500725689405,
+      "grad_norm": 0.6390048265457153,
+      "learning_rate": 2.9836722931398478e-05,
+      "loss": 0.9122,
+      "step": 165
+    },
+    {
+      "epoch": 0.4934687953555878,
+      "grad_norm": 0.7495477199554443,
+      "learning_rate": 2.9814874787568124e-05,
+      "loss": 0.9817,
+      "step": 170
+    },
+    {
+      "epoch": 0.5079825834542816,
+      "grad_norm": 0.8441362977027893,
+      "learning_rate": 2.9791664237522427e-05,
+      "loss": 0.9268,
+      "step": 175
+    },
+    {
+      "epoch": 0.5224963715529753,
+      "grad_norm": 0.7749235033988953,
+      "learning_rate": 2.9767093415751088e-05,
+      "loss": 0.9303,
+      "step": 180
+    },
+    {
+      "epoch": 0.5370101596516691,
+      "grad_norm": 0.8013710379600525,
+      "learning_rate": 2.9741164581837174e-05,
+      "loss": 0.9307,
+      "step": 185
+    },
+    {
+      "epoch": 0.5515239477503628,
+      "grad_norm": 0.7646698355674744,
+      "learning_rate": 2.9713880120249346e-05,
+      "loss": 0.896,
+      "step": 190
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6931925415992737,
+      "learning_rate": 2.9685242540122543e-05,
+      "loss": 0.9353,
+      "step": 195
+    },
+    {
+      "epoch": 0.5805515239477503,
+      "grad_norm": 0.8369060754776001,
+      "learning_rate": 2.9655254475027256e-05,
+      "loss": 0.8669,
+      "step": 200
+    },
+    {
+      "epoch": 0.5950653120464441,
+      "grad_norm": 0.8680087327957153,
+      "learning_rate": 2.9623918682727355e-05,
+      "loss": 0.8872,
+      "step": 205
+    },
+    {
+      "epoch": 0.6095791001451378,
+      "grad_norm": 0.7892455458641052,
+      "learning_rate": 2.9591238044926447e-05,
+      "loss": 0.8725,
+      "step": 210
+    },
+    {
+      "epoch": 0.6240928882438317,
+      "grad_norm": 0.9938645958900452,
+      "learning_rate": 2.9557215567002893e-05,
+      "loss": 0.8602,
+      "step": 215
+    },
+    {
+      "epoch": 0.6386066763425254,
+      "grad_norm": 0.9939874410629272,
+      "learning_rate": 2.9521854377733408e-05,
+      "loss": 0.87,
+      "step": 220
+    },
+    {
+      "epoch": 0.6531204644412192,
+      "grad_norm": 0.8712331652641296,
+      "learning_rate": 2.948515772900536e-05,
+      "loss": 0.7954,
+      "step": 225
+    },
+    {
+      "epoch": 0.6676342525399129,
+      "grad_norm": 0.8074164390563965,
+      "learning_rate": 2.944712899551769e-05,
+      "loss": 0.7585,
+      "step": 230
+    },
+    {
+      "epoch": 0.6821480406386067,
+      "grad_norm": 0.8662286996841431,
+      "learning_rate": 2.9407771674470585e-05,
+      "loss": 0.8194,
+      "step": 235
+    },
+    {
+      "epoch": 0.6966618287373004,
+      "grad_norm": 0.9220862984657288,
+      "learning_rate": 2.936708938524386e-05,
+      "loss": 0.8201,
+      "step": 240
+    },
+    {
+      "epoch": 0.7111756168359942,
+      "grad_norm": 1.0843708515167236,
+      "learning_rate": 2.9325085869064122e-05,
+      "loss": 0.8569,
+      "step": 245
+    },
+    {
+      "epoch": 0.7256894049346879,
+      "grad_norm": 1.1214152574539185,
+      "learning_rate": 2.928176498866071e-05,
+      "loss": 0.8238,
+      "step": 250
+    },
+    {
+      "epoch": 0.7402031930333817,
+      "grad_norm": 0.9205763339996338,
+      "learning_rate": 2.923713072791048e-05,
+      "loss": 0.7594,
+      "step": 255
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.8446714282035828,
+      "learning_rate": 2.919118719147142e-05,
+      "loss": 0.8158,
+      "step": 260
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 1.0141100883483887,
+      "learning_rate": 2.9143938604405202e-05,
+      "loss": 0.7458,
+      "step": 265
+    },
+    {
+      "epoch": 0.783744557329463,
+      "grad_norm": 0.9480670690536499,
+      "learning_rate": 2.9095389311788626e-05,
+      "loss": 0.7803,
+      "step": 270
+    },
+    {
+      "epoch": 0.7982583454281568,
+      "grad_norm": 0.9860946536064148,
+      "learning_rate": 2.904554377831403e-05,
+      "loss": 0.7133,
+      "step": 275
+    },
+    {
+      "epoch": 0.8127721335268505,
+      "grad_norm": 0.9124419689178467,
+      "learning_rate": 2.8994406587878722e-05,
+      "loss": 0.7819,
+      "step": 280
+    },
+    {
+      "epoch": 0.8272859216255443,
+      "grad_norm": 1.042509913444519,
+      "learning_rate": 2.8941982443163426e-05,
+      "loss": 0.7649,
+      "step": 285
+    },
+    {
+      "epoch": 0.841799709724238,
+      "grad_norm": 0.9904513955116272,
+      "learning_rate": 2.888827616519981e-05,
+      "loss": 0.6891,
+      "step": 290
+    },
+    {
+      "epoch": 0.8563134978229318,
+      "grad_norm": 0.9591580629348755,
+      "learning_rate": 2.8833292692927156e-05,
+      "loss": 0.7576,
+      "step": 295
+    },
+    {
+      "epoch": 0.8708272859216255,
+      "grad_norm": 1.0634745359420776,
+      "learning_rate": 2.8777037082738128e-05,
+      "loss": 0.7069,
+      "step": 300
+    },
+    {
+      "epoch": 0.8853410740203193,
+      "grad_norm": 1.0109038352966309,
+      "learning_rate": 2.8719514508013817e-05,
+      "loss": 0.7104,
+      "step": 305
+    },
+    {
+      "epoch": 0.8998548621190131,
+      "grad_norm": 0.9772124290466309,
+      "learning_rate": 2.8660730258647962e-05,
+      "loss": 0.6602,
+      "step": 310
+    },
+    {
+      "epoch": 0.9143686502177069,
+      "grad_norm": 1.0730777978897095,
+      "learning_rate": 2.8600689740560464e-05,
+      "loss": 0.7008,
+      "step": 315
+    },
+    {
+      "epoch": 0.9288824383164006,
+      "grad_norm": 1.1543433666229248,
+      "learning_rate": 2.8539398475200304e-05,
+      "loss": 0.6851,
+      "step": 320
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 1.0269038677215576,
+      "learning_rate": 2.8476862099037712e-05,
+      "loss": 0.7069,
+      "step": 325
+    },
+    {
+      "epoch": 0.9579100145137881,
+      "grad_norm": 1.0626029968261719,
+      "learning_rate": 2.841308636304587e-05,
+      "loss": 0.6872,
+      "step": 330
+    },
+    {
+      "epoch": 0.9724238026124818,
+      "grad_norm": 1.0488626956939697,
+      "learning_rate": 2.8348077132172032e-05,
+      "loss": 0.7008,
+      "step": 335
+    },
+    {
+      "epoch": 0.9869375907111756,
+      "grad_norm": 0.9057682156562805,
+      "learning_rate": 2.8281840384798147e-05,
+      "loss": 0.6706,
+      "step": 340
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.3538520336151123,
+      "learning_rate": 2.821438221219112e-05,
+      "loss": 0.6773,
+      "step": 345
+    },
+    {
+      "epoch": 1.0145137880986939,
+      "grad_norm": 1.0018744468688965,
+      "learning_rate": 2.8145708817942606e-05,
+      "loss": 0.575,
+      "step": 350
+    },
+    {
+      "epoch": 1.0290275761973875,
+      "grad_norm": 0.949688196182251,
+      "learning_rate": 2.8075826517398523e-05,
+      "loss": 0.5709,
+      "step": 355
+    },
+    {
+      "epoch": 1.0435413642960814,
+      "grad_norm": 1.0343447923660278,
+      "learning_rate": 2.8004741737078305e-05,
+      "loss": 0.6286,
+      "step": 360
+    },
+    {
+      "epoch": 1.058055152394775,
+      "grad_norm": 1.0741147994995117,
+      "learning_rate": 2.7932461014083875e-05,
+      "loss": 0.5388,
+      "step": 365
+    },
+    {
+      "epoch": 1.0725689404934688,
+      "grad_norm": 1.1340123414993286,
+      "learning_rate": 2.7858990995498505e-05,
+      "loss": 0.6193,
+      "step": 370
+    },
+    {
+      "epoch": 1.0870827285921625,
+      "grad_norm": 1.3583769798278809,
+      "learning_rate": 2.7784338437775502e-05,
+      "loss": 0.5577,
+      "step": 375
+    },
+    {
+      "epoch": 1.1015965166908563,
+      "grad_norm": 1.2563444375991821,
+      "learning_rate": 2.770851020611692e-05,
+      "loss": 0.6166,
+      "step": 380
+    },
+    {
+      "epoch": 1.11611030478955,
+      "grad_norm": 1.0581562519073486,
+      "learning_rate": 2.7631513273842178e-05,
+      "loss": 0.5249,
+      "step": 385
+    },
+    {
+      "epoch": 1.1306240928882438,
+      "grad_norm": 1.0436842441558838,
+      "learning_rate": 2.7553354721746797e-05,
+      "loss": 0.5518,
+      "step": 390
+    },
+    {
+      "epoch": 1.1451378809869377,
+      "grad_norm": 1.1011406183242798,
+      "learning_rate": 2.747404173745124e-05,
+      "loss": 0.5251,
+      "step": 395
+    },
+    {
+      "epoch": 1.1596516690856313,
+      "grad_norm": 0.9995150566101074,
+      "learning_rate": 2.7393581614739924e-05,
+      "loss": 0.551,
+      "step": 400
+    },
+    {
+      "epoch": 1.1741654571843252,
+      "grad_norm": 1.0945287942886353,
+      "learning_rate": 2.7311981752890446e-05,
+      "loss": 0.5266,
+      "step": 405
+    },
+    {
+      "epoch": 1.1886792452830188,
+      "grad_norm": 1.1231412887573242,
+      "learning_rate": 2.7229249655993164e-05,
+      "loss": 0.5995,
+      "step": 410
+    },
+    {
+      "epoch": 1.2031930333817127,
+      "grad_norm": 1.0969406366348267,
+      "learning_rate": 2.7145392932261085e-05,
+      "loss": 0.5056,
+      "step": 415
+    },
+    {
+      "epoch": 1.2177068214804063,
+      "grad_norm": 1.0740689039230347,
+      "learning_rate": 2.706041929333019e-05,
+      "loss": 0.5104,
+      "step": 420
+    },
+    {
+      "epoch": 1.2322206095791002,
+      "grad_norm": 1.0396003723144531,
+      "learning_rate": 2.697433655355029e-05,
+      "loss": 0.5284,
+      "step": 425
+    },
+    {
+      "epoch": 1.2467343976777938,
+      "grad_norm": 0.9959568381309509,
+      "learning_rate": 2.6887152629266354e-05,
+      "loss": 0.518,
+      "step": 430
+    },
+    {
+      "epoch": 1.2612481857764877,
+      "grad_norm": 1.1106600761413574,
+      "learning_rate": 2.6798875538090565e-05,
+      "loss": 0.4998,
+      "step": 435
+    },
+    {
+      "epoch": 1.2757619738751815,
+      "grad_norm": 1.093956708908081,
+      "learning_rate": 2.6709513398164938e-05,
+      "loss": 0.5424,
+      "step": 440
+    },
+    {
+      "epoch": 1.2902757619738752,
+      "grad_norm": 1.2593542337417603,
+      "learning_rate": 2.6619074427414817e-05,
+      "loss": 0.5224,
+      "step": 445
+    },
+    {
+      "epoch": 1.3047895500725688,
+      "grad_norm": 1.1243525743484497,
+      "learning_rate": 2.6527566942793105e-05,
+      "loss": 0.5213,
+      "step": 450
+    },
+    {
+      "epoch": 1.3193033381712627,
+      "grad_norm": 1.195202112197876,
+      "learning_rate": 2.643499935951542e-05,
+      "loss": 0.5096,
+      "step": 455
+    },
+    {
+      "epoch": 1.3338171262699565,
+      "grad_norm": 1.229646921157837,
+      "learning_rate": 2.6341380190286233e-05,
+      "loss": 0.5101,
+      "step": 460
+    },
+    {
+      "epoch": 1.3483309143686502,
+      "grad_norm": 1.1680142879486084,
+      "learning_rate": 2.6246718044516013e-05,
+      "loss": 0.444,
+      "step": 465
+    },
+    {
+      "epoch": 1.362844702467344,
+      "grad_norm": 1.0221312046051025,
+      "learning_rate": 2.615102162752948e-05,
+      "loss": 0.5211,
+      "step": 470
+    },
+    {
+      "epoch": 1.3773584905660377,
+      "grad_norm": 1.2264982461929321,
+      "learning_rate": 2.605429973976505e-05,
+      "loss": 0.4746,
+      "step": 475
+    },
+    {
+      "epoch": 1.3918722786647315,
+      "grad_norm": 1.2884243726730347,
+      "learning_rate": 2.5956561275965548e-05,
+      "loss": 0.4915,
+      "step": 480
+    },
+    {
+      "epoch": 1.4063860667634254,
+      "grad_norm": 1.1846914291381836,
+      "learning_rate": 2.585781522436019e-05,
+      "loss": 0.468,
+      "step": 485
+    },
+    {
+      "epoch": 1.420899854862119,
+      "grad_norm": 1.09394371509552,
+      "learning_rate": 2.575807066583805e-05,
+      "loss": 0.4837,
+      "step": 490
+    },
+    {
+      "epoch": 1.4354136429608126,
+      "grad_norm": 1.1531438827514648,
+      "learning_rate": 2.565733677311292e-05,
+      "loss": 0.4818,
+      "step": 495
+    },
+    {
+      "epoch": 1.4499274310595065,
+      "grad_norm": 1.1869271993637085,
+      "learning_rate": 2.555562280987982e-05,
+      "loss": 0.4815,
+      "step": 500
+    },
+    {
+      "epoch": 1.4644412191582004,
+      "grad_norm": 0.9555249810218811,
+      "learning_rate": 2.545293812996303e-05,
+      "loss": 0.4516,
+      "step": 505
+    },
+    {
+      "epoch": 1.478955007256894,
+      "grad_norm": 1.2877460718154907,
+      "learning_rate": 2.534929217645594e-05,
+      "loss": 0.4291,
+      "step": 510
+    },
+    {
+      "epoch": 1.4934687953555879,
+      "grad_norm": 1.2242249250411987,
+      "learning_rate": 2.5244694480852624e-05,
+      "loss": 0.4653,
+      "step": 515
+    },
+    {
+      "epoch": 1.5079825834542815,
+      "grad_norm": 1.1914358139038086,
+      "learning_rate": 2.5139154662171316e-05,
+      "loss": 0.4862,
+      "step": 520
+    },
+    {
+      "epoch": 1.5224963715529753,
+      "grad_norm": 1.0356124639511108,
+      "learning_rate": 2.5032682426069803e-05,
+      "loss": 0.4527,
+      "step": 525
+    },
+    {
+      "epoch": 1.5370101596516692,
+      "grad_norm": 1.0836732387542725,
+      "learning_rate": 2.4925287563952894e-05,
+      "loss": 0.4595,
+      "step": 530
+    },
+    {
+      "epoch": 1.5515239477503628,
+      "grad_norm": 1.0991663932800293,
+      "learning_rate": 2.4816979952071974e-05,
+      "loss": 0.4244,
+      "step": 535
+    },
+    {
+      "epoch": 1.5660377358490565,
+      "grad_norm": 1.0540759563446045,
+      "learning_rate": 2.4707769550616762e-05,
+      "loss": 0.4281,
+      "step": 540
+    },
+    {
+      "epoch": 1.5805515239477503,
+      "grad_norm": 1.142854928970337,
+      "learning_rate": 2.4597666402799363e-05,
+      "loss": 0.4506,
+      "step": 545
+    },
+    {
+      "epoch": 1.5950653120464442,
+      "grad_norm": 1.1945523023605347,
+      "learning_rate": 2.4486680633930658e-05,
+      "loss": 0.4435,
+      "step": 550
+    },
+    {
+      "epoch": 1.6095791001451378,
+      "grad_norm": 1.200697898864746,
+      "learning_rate": 2.4374822450489173e-05,
+      "loss": 0.4331,
+      "step": 555
+    },
+    {
+      "epoch": 1.6240928882438317,
+      "grad_norm": 1.0646142959594727,
+      "learning_rate": 2.4262102139182453e-05,
+      "loss": 0.4211,
+      "step": 560
+    },
+    {
+      "epoch": 1.6386066763425253,
+      "grad_norm": 1.0593570470809937,
+      "learning_rate": 2.414853006600109e-05,
+      "loss": 0.397,
+      "step": 565
+    },
+    {
+      "epoch": 1.6531204644412192,
+      "grad_norm": 1.0598646402359009,
+      "learning_rate": 2.403411667526546e-05,
+      "loss": 0.408,
+      "step": 570
+    },
+    {
+      "epoch": 1.667634252539913,
+      "grad_norm": 1.0258424282073975,
+      "learning_rate": 2.3918872488665195e-05,
+      "loss": 0.374,
+      "step": 575
+    },
+    {
+      "epoch": 1.6821480406386067,
+      "grad_norm": 1.2203288078308105,
+      "learning_rate": 2.380280810429164e-05,
+      "loss": 0.4155,
+      "step": 580
+    },
+    {
+      "epoch": 1.6966618287373003,
+      "grad_norm": 1.2701225280761719,
+      "learning_rate": 2.3685934195663184e-05,
+      "loss": 0.3916,
+      "step": 585
+    },
+    {
+      "epoch": 1.7111756168359942,
+      "grad_norm": 1.2071703672409058,
+      "learning_rate": 2.3568261510743743e-05,
+      "loss": 0.3882,
+      "step": 590
+    },
+    {
+      "epoch": 1.725689404934688,
+      "grad_norm": 1.0981299877166748,
+      "learning_rate": 2.344980087095433e-05,
+      "loss": 0.4114,
+      "step": 595
+    },
+    {
+      "epoch": 1.7402031930333817,
+      "grad_norm": 1.2279996871948242,
+      "learning_rate": 2.3330563170177896e-05,
+      "loss": 0.4341,
+      "step": 600
+    },
+    {
+      "epoch": 1.7547169811320755,
+      "grad_norm": 1.215522050857544,
+      "learning_rate": 2.3210559373757515e-05,
+      "loss": 0.4438,
+      "step": 605
+    },
+    {
+      "epoch": 1.7692307692307692,
+      "grad_norm": 1.167236566543579,
+      "learning_rate": 2.308980051748799e-05,
+      "loss": 0.3922,
+      "step": 610
+    },
+    {
+      "epoch": 1.783744557329463,
+      "grad_norm": 1.12175714969635,
+      "learning_rate": 2.296829770660096e-05,
+      "loss": 0.3864,
+      "step": 615
+    },
+    {
+      "epoch": 1.7982583454281569,
+      "grad_norm": 1.1657911539077759,
+      "learning_rate": 2.2846062114743665e-05,
+      "loss": 0.3834,
+      "step": 620
+    },
+    {
+      "epoch": 1.8127721335268505,
+      "grad_norm": 1.0777708292007446,
+      "learning_rate": 2.272310498295138e-05,
+      "loss": 0.388,
+      "step": 625
+    },
+    {
+      "epoch": 1.8272859216255442,
+      "grad_norm": 1.0510711669921875,
+      "learning_rate": 2.259943761861367e-05,
+      "loss": 0.3934,
+      "step": 630
+    },
+    {
+      "epoch": 1.841799709724238,
+      "grad_norm": 1.141183853149414,
+      "learning_rate": 2.247507139443454e-05,
+      "loss": 0.3804,
+      "step": 635
+    },
+    {
+      "epoch": 1.8563134978229319,
+      "grad_norm": 1.1873807907104492,
+      "learning_rate": 2.2350017747386573e-05,
+      "loss": 0.3764,
+      "step": 640
+    },
+    {
+      "epoch": 1.8708272859216255,
+      "grad_norm": 1.1094430685043335,
+      "learning_rate": 2.2224288177659174e-05,
+      "loss": 0.3557,
+      "step": 645
+    },
+    {
+      "epoch": 1.8853410740203191,
+      "grad_norm": 1.0801291465759277,
+      "learning_rate": 2.2097894247600965e-05,
+      "loss": 0.333,
+      "step": 650
+    },
+    {
+      "epoch": 1.899854862119013,
+      "grad_norm": 1.171615719795227,
+      "learning_rate": 2.197084758065653e-05,
+      "loss": 0.3844,
+      "step": 655
+    },
+    {
+      "epoch": 1.9143686502177069,
+      "grad_norm": 1.1228846311569214,
+      "learning_rate": 2.1843159860297445e-05,
+      "loss": 0.3729,
+      "step": 660
+    },
+    {
+      "epoch": 1.9288824383164007,
+      "grad_norm": 1.0446889400482178,
+      "learning_rate": 2.171484282894789e-05,
+      "loss": 0.3764,
+      "step": 665
+    },
+    {
+      "epoch": 1.9433962264150944,
+      "grad_norm": 1.055737853050232,
+      "learning_rate": 2.158590828690477e-05,
+      "loss": 0.3313,
+      "step": 670
+    },
+    {
+      "epoch": 1.957910014513788,
+      "grad_norm": 1.2090764045715332,
+      "learning_rate": 2.1456368091252526e-05,
+      "loss": 0.3725,
+      "step": 675
+    },
+    {
+      "epoch": 1.9724238026124818,
+      "grad_norm": 1.0745823383331299,
+      "learning_rate": 2.1326234154772752e-05,
+      "loss": 0.3492,
+      "step": 680
+    },
+    {
+      "epoch": 1.9869375907111757,
+      "grad_norm": 1.2478907108306885,
+      "learning_rate": 2.1195518444848673e-05,
+      "loss": 0.3344,
+      "step": 685
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.450290560722351,
+      "learning_rate": 2.1064232982364572e-05,
+      "loss": 0.3748,
+      "step": 690
+    },
+    {
+      "epoch": 2.014513788098694,
+      "grad_norm": 1.1875301599502563,
+      "learning_rate": 2.093238984060037e-05,
+      "loss": 0.2611,
+      "step": 695
+    },
+    {
+      "epoch": 2.0290275761973877,
+      "grad_norm": 1.0669788122177124,
+      "learning_rate": 2.0800001144121284e-05,
+      "loss": 0.267,
+      "step": 700
+    },
+    {
+      "epoch": 2.043541364296081,
+      "grad_norm": 1.1223045587539673,
+      "learning_rate": 2.0667079067662882e-05,
+      "loss": 0.2914,
+      "step": 705
+    },
+    {
+      "epoch": 2.058055152394775,
+      "grad_norm": 0.9819235801696777,
+      "learning_rate": 2.0533635835011448e-05,
+      "loss": 0.3139,
+      "step": 710
+    },
+    {
+      "epoch": 2.072568940493469,
+      "grad_norm": 1.3752785921096802,
+      "learning_rate": 2.0399683717879837e-05,
+      "loss": 0.2765,
+      "step": 715
+    },
+    {
+      "epoch": 2.0870827285921627,
+      "grad_norm": 1.0086604356765747,
+      "learning_rate": 2.0265235034778983e-05,
+      "loss": 0.3018,
+      "step": 720
+    },
+    {
+      "epoch": 2.101596516690856,
+      "grad_norm": 1.1323165893554688,
+      "learning_rate": 2.0130302149885033e-05,
+      "loss": 0.3085,
+      "step": 725
+    },
+    {
+      "epoch": 2.11611030478955,
+      "grad_norm": 0.9608115553855896,
+      "learning_rate": 1.9994897471902325e-05,
+      "loss": 0.2563,
+      "step": 730
+    },
+    {
+      "epoch": 2.130624092888244,
+      "grad_norm": 1.293845534324646,
+      "learning_rate": 1.9859033452922253e-05,
+      "loss": 0.2519,
+      "step": 735
+    },
+    {
+      "epoch": 2.1451378809869377,
+      "grad_norm": 1.0907076597213745,
+      "learning_rate": 1.972272258727815e-05,
+      "loss": 0.2778,
+      "step": 740
+    },
+    {
+      "epoch": 2.1596516690856316,
+      "grad_norm": 1.265874981880188,
+      "learning_rate": 1.9585977410396295e-05,
+      "loss": 0.2577,
+      "step": 745
+    },
+    {
+      "epoch": 2.174165457184325,
+      "grad_norm": 1.1681596040725708,
+      "learning_rate": 1.9448810497643095e-05,
+      "loss": 0.2465,
+      "step": 750
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 1.2213428020477295,
+      "learning_rate": 1.9311234463168666e-05,
+      "loss": 0.259,
+      "step": 755
+    },
+    {
+      "epoch": 2.2031930333817127,
+      "grad_norm": 1.3326306343078613,
+      "learning_rate": 1.9173261958746793e-05,
+      "loss": 0.248,
+      "step": 760
+    },
+    {
+      "epoch": 2.2177068214804065,
+      "grad_norm": 1.1038895845413208,
+      "learning_rate": 1.9034905672611445e-05,
+      "loss": 0.2512,
+      "step": 765
+    },
+    {
+      "epoch": 2.2322206095791,
+      "grad_norm": 1.180465817451477,
+      "learning_rate": 1.889617832828995e-05,
+      "loss": 0.2609,
+      "step": 770
+    },
+    {
+      "epoch": 2.246734397677794,
+      "grad_norm": 1.1591984033584595,
+      "learning_rate": 1.8757092683432903e-05,
+      "loss": 0.2935,
+      "step": 775
+    },
+    {
+      "epoch": 2.2612481857764877,
+      "grad_norm": 1.2751474380493164,
+      "learning_rate": 1.8617661528640946e-05,
+      "loss": 0.2471,
+      "step": 780
+    },
+    {
+      "epoch": 2.2757619738751815,
+      "grad_norm": 1.1469013690948486,
+      "learning_rate": 1.8477897686288523e-05,
+      "loss": 0.2455,
+      "step": 785
+    },
+    {
+      "epoch": 2.2902757619738754,
+      "grad_norm": 1.1740764379501343,
+      "learning_rate": 1.8337814009344716e-05,
+      "loss": 0.2753,
+      "step": 790
+    },
+    {
+      "epoch": 2.304789550072569,
+      "grad_norm": 1.2813502550125122,
+      "learning_rate": 1.819742338019124e-05,
+      "loss": 0.2476,
+      "step": 795
+    },
+    {
+      "epoch": 2.3193033381712627,
+      "grad_norm": 1.2389832735061646,
+      "learning_rate": 1.8056738709437775e-05,
+      "loss": 0.2408,
+      "step": 800
+    },
+    {
+      "epoch": 2.3338171262699565,
+      "grad_norm": 1.0330638885498047,
+      "learning_rate": 1.7915772934734653e-05,
+      "loss": 0.2375,
+      "step": 805
+    },
+    {
+      "epoch": 2.3483309143686504,
+      "grad_norm": 1.3108922243118286,
+      "learning_rate": 1.7774539019583108e-05,
+      "loss": 0.2407,
+      "step": 810
+    },
+    {
+      "epoch": 2.362844702467344,
+      "grad_norm": 1.1805189847946167,
+      "learning_rate": 1.763304995214313e-05,
+      "loss": 0.2445,
+      "step": 815
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 1.2893935441970825,
+      "learning_rate": 1.7491318744039012e-05,
+      "loss": 0.2551,
+      "step": 820
+    },
+    {
+      "epoch": 2.3918722786647315,
+      "grad_norm": 1.0978128910064697,
+      "learning_rate": 1.734935842916282e-05,
+      "loss": 0.2377,
+      "step": 825
+    },
+    {
+      "epoch": 2.4063860667634254,
+      "grad_norm": 1.1712990999221802,
+      "learning_rate": 1.720718206247574e-05,
+      "loss": 0.2334,
+      "step": 830
+    },
+    {
+      "epoch": 2.4208998548621192,
+      "grad_norm": 1.3427225351333618,
+      "learning_rate": 1.7064802718807526e-05,
+      "loss": 0.2321,
+      "step": 835
+    },
+    {
+      "epoch": 2.4354136429608126,
+      "grad_norm": 1.1522012948989868,
+      "learning_rate": 1.69222334916541e-05,
+      "loss": 0.2443,
+      "step": 840
+    },
+    {
+      "epoch": 2.4499274310595065,
+      "grad_norm": 1.1092753410339355,
+      "learning_rate": 1.6779487491973472e-05,
+      "loss": 0.2327,
+      "step": 845
+    },
+    {
+      "epoch": 2.4644412191582004,
+      "grad_norm": 1.0141578912734985,
+      "learning_rate": 1.6636577846980014e-05,
+      "loss": 0.2435,
+      "step": 850
+    },
+    {
+      "epoch": 2.478955007256894,
+      "grad_norm": 1.3944146633148193,
+      "learning_rate": 1.6493517698937253e-05,
+      "loss": 0.2231,
+      "step": 855
+    },
+    {
+      "epoch": 2.4934687953555876,
+      "grad_norm": 1.0414365530014038,
+      "learning_rate": 1.6350320203949274e-05,
+      "loss": 0.2245,
+      "step": 860
+    },
+    {
+      "epoch": 2.5079825834542815,
+      "grad_norm": 1.1255793571472168,
+      "learning_rate": 1.6206998530750893e-05,
+      "loss": 0.2081,
+      "step": 865
+    },
+    {
+      "epoch": 2.5224963715529753,
+      "grad_norm": 1.2589868307113647,
+      "learning_rate": 1.6063565859496577e-05,
+      "loss": 0.232,
+      "step": 870
+    },
+    {
+      "epoch": 2.537010159651669,
+      "grad_norm": 1.2401530742645264,
+      "learning_rate": 1.5920035380548433e-05,
+      "loss": 0.216,
+      "step": 875
+    },
+    {
+      "epoch": 2.551523947750363,
+      "grad_norm": 1.092507243156433,
+      "learning_rate": 1.5776420293263144e-05,
+      "loss": 0.2228,
+      "step": 880
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 1.1913965940475464,
+      "learning_rate": 1.5632733804778154e-05,
+      "loss": 0.2006,
+      "step": 885
+    },
+    {
+      "epoch": 2.5805515239477503,
+      "grad_norm": 1.0700242519378662,
+      "learning_rate": 1.54889891287971e-05,
+      "loss": 0.2449,
+      "step": 890
+    },
+    {
+      "epoch": 2.595065312046444,
+      "grad_norm": 1.0306884050369263,
+      "learning_rate": 1.534519948437468e-05,
+      "loss": 0.1815,
+      "step": 895
+    },
+    {
+      "epoch": 2.6095791001451376,
+      "grad_norm": 1.108221173286438,
+      "learning_rate": 1.5201378094700963e-05,
+      "loss": 0.2089,
+      "step": 900
+    },
+    {
+      "epoch": 2.6240928882438315,
+      "grad_norm": 1.1428337097167969,
+      "learning_rate": 1.5057538185885385e-05,
+      "loss": 0.1899,
+      "step": 905
+    },
+    {
+      "epoch": 2.6386066763425253,
+      "grad_norm": 1.193100094795227,
+      "learning_rate": 1.4913692985740438e-05,
+      "loss": 0.2571,
+      "step": 910
+    },
+    {
+      "epoch": 2.653120464441219,
+      "grad_norm": 1.167102575302124,
+      "learning_rate": 1.4769855722565214e-05,
+      "loss": 0.25,
+      "step": 915
+    },
+    {
+      "epoch": 2.667634252539913,
+      "grad_norm": 1.152069091796875,
+      "learning_rate": 1.462603962392891e-05,
+      "loss": 0.2327,
+      "step": 920
+    },
+    {
+      "epoch": 2.682148040638607,
+      "grad_norm": 1.0608423948287964,
+      "learning_rate": 1.4482257915454383e-05,
+      "loss": 0.2055,
+      "step": 925
+    },
+    {
+      "epoch": 2.6966618287373003,
+      "grad_norm": 1.1504971981048584,
+      "learning_rate": 1.43385238196019e-05,
+      "loss": 0.2048,
+      "step": 930
+    },
+    {
+      "epoch": 2.711175616835994,
+      "grad_norm": 1.1371382474899292,
+      "learning_rate": 1.4194850554453175e-05,
+      "loss": 0.2271,
+      "step": 935
+    },
+    {
+      "epoch": 2.725689404934688,
+      "grad_norm": 1.0657782554626465,
+      "learning_rate": 1.4051251332495819e-05,
+      "loss": 0.2092,
+      "step": 940
+    },
+    {
+      "epoch": 2.7402031930333814,
+      "grad_norm": 1.3242257833480835,
+      "learning_rate": 1.3907739359408259e-05,
+      "loss": 0.2295,
+      "step": 945
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 1.1220176219940186,
+      "learning_rate": 1.3764327832845344e-05,
+      "loss": 0.1953,
+      "step": 950
+    },
+    {
+      "epoch": 2.769230769230769,
+      "grad_norm": 1.0856845378875732,
+      "learning_rate": 1.3621029941224666e-05,
+      "loss": 0.2045,
+      "step": 955
+    },
+    {
+      "epoch": 2.783744557329463,
+      "grad_norm": 1.239922046661377,
+      "learning_rate": 1.3477858862513698e-05,
+      "loss": 0.203,
+      "step": 960
+    },
+    {
+      "epoch": 2.798258345428157,
+      "grad_norm": 1.0360215902328491,
+      "learning_rate": 1.3334827763017952e-05,
+      "loss": 0.2098,
+      "step": 965
+    },
+    {
+      "epoch": 2.8127721335268507,
+      "grad_norm": 1.1175003051757812,
+      "learning_rate": 1.3191949796170156e-05,
+      "loss": 0.1812,
+      "step": 970
+    },
+    {
+      "epoch": 2.827285921625544,
+      "grad_norm": 1.1741539239883423,
+      "learning_rate": 1.3049238101320652e-05,
+      "loss": 0.2123,
+      "step": 975
+    },
+    {
+      "epoch": 2.841799709724238,
+      "grad_norm": 1.0360662937164307,
+      "learning_rate": 1.2906705802529064e-05,
+      "loss": 0.1724,
+      "step": 980
+    },
+    {
+      "epoch": 2.856313497822932,
+      "grad_norm": 1.0688581466674805,
+      "learning_rate": 1.2764366007357382e-05,
+      "loss": 0.174,
+      "step": 985
+    },
+    {
+      "epoch": 2.8708272859216253,
+      "grad_norm": 1.1229158639907837,
+      "learning_rate": 1.2622231805664592e-05,
+      "loss": 0.2262,
+      "step": 990
+    },
+    {
+      "epoch": 2.885341074020319,
+      "grad_norm": 1.1505873203277588,
+      "learning_rate": 1.2480316268402859e-05,
+      "loss": 0.1717,
+      "step": 995
+    },
+    {
+      "epoch": 2.899854862119013,
+      "grad_norm": 1.1841315031051636,
+      "learning_rate": 1.2338632446415526e-05,
+      "loss": 0.1993,
+      "step": 1000
+    },
+    {
+      "epoch": 2.914368650217707,
+      "grad_norm": 1.1179850101470947,
+      "learning_rate": 1.2197193369236929e-05,
+      "loss": 0.1783,
+      "step": 1005
+    },
+    {
+      "epoch": 2.9288824383164007,
+      "grad_norm": 1.2388747930526733,
+      "learning_rate": 1.2056012043894166e-05,
+      "loss": 0.1867,
+      "step": 1010
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 1.0603691339492798,
+      "learning_rate": 1.1915101453710953e-05,
+      "loss": 0.1724,
+      "step": 1015
+    },
+    {
+      "epoch": 2.957910014513788,
+      "grad_norm": 1.209848165512085,
+      "learning_rate": 1.1774474557113642e-05,
+      "loss": 0.2029,
+      "step": 1020
+    },
+    {
+      "epoch": 2.972423802612482,
+      "grad_norm": 1.1248085498809814,
+      "learning_rate": 1.163414428643954e-05,
+      "loss": 0.2085,
+      "step": 1025
+    },
+    {
+      "epoch": 2.9869375907111757,
+      "grad_norm": 1.0725529193878174,
+      "learning_rate": 1.1494123546747638e-05,
+      "loss": 0.1946,
+      "step": 1030
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.5596674680709839,
+      "learning_rate": 1.1354425214631804e-05,
+      "loss": 0.1845,
+      "step": 1035
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1725,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5208195517914808e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

40_128_e5_3e-5/checkpoint-1035/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb03dfd1a09e76ed2bbbbab487b8a4cf3a040b6e2f6cfe9b561e1a7b4b2a3466
+size 7736

40_128_e5_3e-5/checkpoint-1035/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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 param_shapes.items():
+
+        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_file, 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

40_128_e5_3e-5/checkpoint-1380/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+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

40_128_e5_3e-5/checkpoint-1380/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

40_128_e5_3e-5/checkpoint-1380/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df753eac0a54d0cc3b7ed5e4e52da9fe274f9e7cdb0f0204c3f56c361d8aa36a
+size 791751704

40_128_e5_3e-5/checkpoint-1380/latest

@@ -0,0 +1 @@
+global_step1380

40_128_e5_3e-5/checkpoint-1380/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93dfd14a1f5237c87218676c739b27893c872b566b4af920e3b8ac6afcc25fda
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0afc549d186978268ebcfe65fc98cc5dc05905afba0f988a5df565a2b20b0ebb
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36049a8a07b14c750aca48508807a87b18dfc3a8294e5e87f4aab75a9eb5b17d
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f63535f6c8ae26a05988e45024fc6186dbf609a3dc265a0f4a6edcd0690368a
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c79d05ee83969d2368e6df38ce22b3c34481709c49a8803ca14ae3212b5bd8bd
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccd64f5368a079c430fd228e8abca0bb111b9190d35fd530441c70f12e57caaa
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbf673a0e823564555c840705c27756e7e66abb0580c7a120d27873a3c394d8d
+size 15920

40_128_e5_3e-5/checkpoint-1380/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82cbebb1e7bff08db1e6c9c47dc2a147c71053087a1c03f9014c9fd20f46ef07
+size 15920

40_128_e5_3e-5/checkpoint-1380/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dbc868d4607526550b8e8f3aa0648d38b107624a30d4044adf8d97475578d0e
+size 1064

40_128_e5_3e-5/checkpoint-1380/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

40_128_e5_3e-5/checkpoint-1380/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

40_128_e5_3e-5/checkpoint-1380/trainer_state.json

@@ -0,0 +1,1966 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01451378809869376,
+      "grad_norm": 1.1063576936721802,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 1.3635,
+      "step": 5
+    },
+    {
+      "epoch": 0.02902757619738752,
+      "grad_norm": 0.8837594389915466,
+      "learning_rate": 3.103448275862069e-06,
+      "loss": 1.3281,
+      "step": 10
+    },
+    {
+      "epoch": 0.04354136429608128,
+      "grad_norm": 0.6033729314804077,
+      "learning_rate": 4.827586206896552e-06,
+      "loss": 1.37,
+      "step": 15
+    },
+    {
+      "epoch": 0.05805515239477504,
+      "grad_norm": 0.6915555000305176,
+      "learning_rate": 6.551724137931035e-06,
+      "loss": 1.2765,
+      "step": 20
+    },
+    {
+      "epoch": 0.07256894049346879,
+      "grad_norm": 0.7305431365966797,
+      "learning_rate": 8.275862068965517e-06,
+      "loss": 1.2649,
+      "step": 25
+    },
+    {
+      "epoch": 0.08708272859216255,
+      "grad_norm": 0.47839707136154175,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 1.2497,
+      "step": 30
+    },
+    {
+      "epoch": 0.10159651669085631,
+      "grad_norm": 0.5838127136230469,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 1.2576,
+      "step": 35
+    },
+    {
+      "epoch": 0.11611030478955008,
+      "grad_norm": 0.45399269461631775,
+      "learning_rate": 1.3448275862068966e-05,
+      "loss": 1.2541,
+      "step": 40
+    },
+    {
+      "epoch": 0.13062409288824384,
+      "grad_norm": 0.45961660146713257,
+      "learning_rate": 1.517241379310345e-05,
+      "loss": 1.1878,
+      "step": 45
+    },
+    {
+      "epoch": 0.14513788098693758,
+      "grad_norm": 0.5068319439888,
+      "learning_rate": 1.6896551724137932e-05,
+      "loss": 1.2369,
+      "step": 50
+    },
+    {
+      "epoch": 0.15965166908563136,
+      "grad_norm": 0.44543999433517456,
+      "learning_rate": 1.8620689655172415e-05,
+      "loss": 1.2221,
+      "step": 55
+    },
+    {
+      "epoch": 0.1741654571843251,
+      "grad_norm": 0.47043296694755554,
+      "learning_rate": 2.0344827586206894e-05,
+      "loss": 1.2211,
+      "step": 60
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.5043661594390869,
+      "learning_rate": 2.206896551724138e-05,
+      "loss": 1.1938,
+      "step": 65
+    },
+    {
+      "epoch": 0.20319303338171263,
+      "grad_norm": 0.4813358783721924,
+      "learning_rate": 2.3793103448275862e-05,
+      "loss": 1.2066,
+      "step": 70
+    },
+    {
+      "epoch": 0.21770682148040638,
+      "grad_norm": 0.44516244530677795,
+      "learning_rate": 2.5517241379310345e-05,
+      "loss": 1.175,
+      "step": 75
+    },
+    {
+      "epoch": 0.23222060957910015,
+      "grad_norm": 0.5195889472961426,
+      "learning_rate": 2.7241379310344827e-05,
+      "loss": 1.1696,
+      "step": 80
+    },
+    {
+      "epoch": 0.2467343976777939,
+      "grad_norm": 0.45658233761787415,
+      "learning_rate": 2.8965517241379313e-05,
+      "loss": 1.1516,
+      "step": 85
+    },
+    {
+      "epoch": 0.2612481857764877,
+      "grad_norm": 0.4927058517932892,
+      "learning_rate": 2.999988964483289e-05,
+      "loss": 1.1275,
+      "step": 90
+    },
+    {
+      "epoch": 0.2757619738751814,
+      "grad_norm": 0.5645100474357605,
+      "learning_rate": 2.9998648167850816e-05,
+      "loss": 1.1119,
+      "step": 95
+    },
+    {
+      "epoch": 0.29027576197387517,
+      "grad_norm": 0.5247628688812256,
+      "learning_rate": 2.9996027384477114e-05,
+      "loss": 1.1642,
+      "step": 100
+    },
+    {
+      "epoch": 0.3047895500725689,
+      "grad_norm": 0.562659740447998,
+      "learning_rate": 2.999202753572439e-05,
+      "loss": 1.1142,
+      "step": 105
+    },
+    {
+      "epoch": 0.3193033381712627,
+      "grad_norm": 0.5908432602882385,
+      "learning_rate": 2.998664898942692e-05,
+      "loss": 1.0488,
+      "step": 110
+    },
+    {
+      "epoch": 0.33381712626995647,
+      "grad_norm": 0.6552732586860657,
+      "learning_rate": 2.997989224020684e-05,
+      "loss": 1.1,
+      "step": 115
+    },
+    {
+      "epoch": 0.3483309143686502,
+      "grad_norm": 0.5694405436515808,
+      "learning_rate": 2.9971757909428642e-05,
+      "loss": 1.0119,
+      "step": 120
+    },
+    {
+      "epoch": 0.36284470246734396,
+      "grad_norm": 0.6327081918716431,
+      "learning_rate": 2.996224674514203e-05,
+      "loss": 1.0156,
+      "step": 125
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.60078364610672,
+      "learning_rate": 2.995135962201315e-05,
+      "loss": 1.0178,
+      "step": 130
+    },
+    {
+      "epoch": 0.3918722786647315,
+      "grad_norm": 0.7008715867996216,
+      "learning_rate": 2.993909754124413e-05,
+      "loss": 0.9871,
+      "step": 135
+    },
+    {
+      "epoch": 0.40638606676342526,
+      "grad_norm": 0.6301591396331787,
+      "learning_rate": 2.9925461630481022e-05,
+      "loss": 1.0024,
+      "step": 140
+    },
+    {
+      "epoch": 0.420899854862119,
+      "grad_norm": 0.6791720390319824,
+      "learning_rate": 2.9910453143710096e-05,
+      "loss": 0.9605,
+      "step": 145
+    },
+    {
+      "epoch": 0.43541364296081275,
+      "grad_norm": 0.650157630443573,
+      "learning_rate": 2.9894073461142523e-05,
+      "loss": 1.0019,
+      "step": 150
+    },
+    {
+      "epoch": 0.44992743105950656,
+      "grad_norm": 0.7557404637336731,
+      "learning_rate": 2.9876324089087442e-05,
+      "loss": 0.9906,
+      "step": 155
+    },
+    {
+      "epoch": 0.4644412191582003,
+      "grad_norm": 0.74793940782547,
+      "learning_rate": 2.9857206659813447e-05,
+      "loss": 1.0025,
+      "step": 160
+    },
+    {
+      "epoch": 0.47895500725689405,
+      "grad_norm": 0.6390048265457153,
+      "learning_rate": 2.9836722931398478e-05,
+      "loss": 0.9122,
+      "step": 165
+    },
+    {
+      "epoch": 0.4934687953555878,
+      "grad_norm": 0.7495477199554443,
+      "learning_rate": 2.9814874787568124e-05,
+      "loss": 0.9817,
+      "step": 170
+    },
+    {
+      "epoch": 0.5079825834542816,
+      "grad_norm": 0.8441362977027893,
+      "learning_rate": 2.9791664237522427e-05,
+      "loss": 0.9268,
+      "step": 175
+    },
+    {
+      "epoch": 0.5224963715529753,
+      "grad_norm": 0.7749235033988953,
+      "learning_rate": 2.9767093415751088e-05,
+      "loss": 0.9303,
+      "step": 180
+    },
+    {
+      "epoch": 0.5370101596516691,
+      "grad_norm": 0.8013710379600525,
+      "learning_rate": 2.9741164581837174e-05,
+      "loss": 0.9307,
+      "step": 185
+    },
+    {
+      "epoch": 0.5515239477503628,
+      "grad_norm": 0.7646698355674744,
+      "learning_rate": 2.9713880120249346e-05,
+      "loss": 0.896,
+      "step": 190
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6931925415992737,
+      "learning_rate": 2.9685242540122543e-05,
+      "loss": 0.9353,
+      "step": 195
+    },
+    {
+      "epoch": 0.5805515239477503,
+      "grad_norm": 0.8369060754776001,
+      "learning_rate": 2.9655254475027256e-05,
+      "loss": 0.8669,
+      "step": 200
+    },
+    {
+      "epoch": 0.5950653120464441,
+      "grad_norm": 0.8680087327957153,
+      "learning_rate": 2.9623918682727355e-05,
+      "loss": 0.8872,
+      "step": 205
+    },
+    {
+      "epoch": 0.6095791001451378,
+      "grad_norm": 0.7892455458641052,
+      "learning_rate": 2.9591238044926447e-05,
+      "loss": 0.8725,
+      "step": 210
+    },
+    {
+      "epoch": 0.6240928882438317,
+      "grad_norm": 0.9938645958900452,
+      "learning_rate": 2.9557215567002893e-05,
+      "loss": 0.8602,
+      "step": 215
+    },
+    {
+      "epoch": 0.6386066763425254,
+      "grad_norm": 0.9939874410629272,
+      "learning_rate": 2.9521854377733408e-05,
+      "loss": 0.87,
+      "step": 220
+    },
+    {
+      "epoch": 0.6531204644412192,
+      "grad_norm": 0.8712331652641296,
+      "learning_rate": 2.948515772900536e-05,
+      "loss": 0.7954,
+      "step": 225
+    },
+    {
+      "epoch": 0.6676342525399129,
+      "grad_norm": 0.8074164390563965,
+      "learning_rate": 2.944712899551769e-05,
+      "loss": 0.7585,
+      "step": 230
+    },
+    {
+      "epoch": 0.6821480406386067,
+      "grad_norm": 0.8662286996841431,
+      "learning_rate": 2.9407771674470585e-05,
+      "loss": 0.8194,
+      "step": 235
+    },
+    {
+      "epoch": 0.6966618287373004,
+      "grad_norm": 0.9220862984657288,
+      "learning_rate": 2.936708938524386e-05,
+      "loss": 0.8201,
+      "step": 240
+    },
+    {
+      "epoch": 0.7111756168359942,
+      "grad_norm": 1.0843708515167236,
+      "learning_rate": 2.9325085869064122e-05,
+      "loss": 0.8569,
+      "step": 245
+    },
+    {
+      "epoch": 0.7256894049346879,
+      "grad_norm": 1.1214152574539185,
+      "learning_rate": 2.928176498866071e-05,
+      "loss": 0.8238,
+      "step": 250
+    },
+    {
+      "epoch": 0.7402031930333817,
+      "grad_norm": 0.9205763339996338,
+      "learning_rate": 2.923713072791048e-05,
+      "loss": 0.7594,
+      "step": 255
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.8446714282035828,
+      "learning_rate": 2.919118719147142e-05,
+      "loss": 0.8158,
+      "step": 260
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 1.0141100883483887,
+      "learning_rate": 2.9143938604405202e-05,
+      "loss": 0.7458,
+      "step": 265
+    },
+    {
+      "epoch": 0.783744557329463,
+      "grad_norm": 0.9480670690536499,
+      "learning_rate": 2.9095389311788626e-05,
+      "loss": 0.7803,
+      "step": 270
+    },
+    {
+      "epoch": 0.7982583454281568,
+      "grad_norm": 0.9860946536064148,
+      "learning_rate": 2.904554377831403e-05,
+      "loss": 0.7133,
+      "step": 275
+    },
+    {
+      "epoch": 0.8127721335268505,
+      "grad_norm": 0.9124419689178467,
+      "learning_rate": 2.8994406587878722e-05,
+      "loss": 0.7819,
+      "step": 280
+    },
+    {
+      "epoch": 0.8272859216255443,
+      "grad_norm": 1.042509913444519,
+      "learning_rate": 2.8941982443163426e-05,
+      "loss": 0.7649,
+      "step": 285
+    },
+    {
+      "epoch": 0.841799709724238,
+      "grad_norm": 0.9904513955116272,
+      "learning_rate": 2.888827616519981e-05,
+      "loss": 0.6891,
+      "step": 290
+    },
+    {
+      "epoch": 0.8563134978229318,
+      "grad_norm": 0.9591580629348755,
+      "learning_rate": 2.8833292692927156e-05,
+      "loss": 0.7576,
+      "step": 295
+    },
+    {
+      "epoch": 0.8708272859216255,
+      "grad_norm": 1.0634745359420776,
+      "learning_rate": 2.8777037082738128e-05,
+      "loss": 0.7069,
+      "step": 300
+    },
+    {
+      "epoch": 0.8853410740203193,
+      "grad_norm": 1.0109038352966309,
+      "learning_rate": 2.8719514508013817e-05,
+      "loss": 0.7104,
+      "step": 305
+    },
+    {
+      "epoch": 0.8998548621190131,
+      "grad_norm": 0.9772124290466309,
+      "learning_rate": 2.8660730258647962e-05,
+      "loss": 0.6602,
+      "step": 310
+    },
+    {
+      "epoch": 0.9143686502177069,
+      "grad_norm": 1.0730777978897095,
+      "learning_rate": 2.8600689740560464e-05,
+      "loss": 0.7008,
+      "step": 315
+    },
+    {
+      "epoch": 0.9288824383164006,
+      "grad_norm": 1.1543433666229248,
+      "learning_rate": 2.8539398475200304e-05,
+      "loss": 0.6851,
+      "step": 320
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 1.0269038677215576,
+      "learning_rate": 2.8476862099037712e-05,
+      "loss": 0.7069,
+      "step": 325
+    },
+    {
+      "epoch": 0.9579100145137881,
+      "grad_norm": 1.0626029968261719,
+      "learning_rate": 2.841308636304587e-05,
+      "loss": 0.6872,
+      "step": 330
+    },
+    {
+      "epoch": 0.9724238026124818,
+      "grad_norm": 1.0488626956939697,
+      "learning_rate": 2.8348077132172032e-05,
+      "loss": 0.7008,
+      "step": 335
+    },
+    {
+      "epoch": 0.9869375907111756,
+      "grad_norm": 0.9057682156562805,
+      "learning_rate": 2.8281840384798147e-05,
+      "loss": 0.6706,
+      "step": 340
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.3538520336151123,
+      "learning_rate": 2.821438221219112e-05,
+      "loss": 0.6773,
+      "step": 345
+    },
+    {
+      "epoch": 1.0145137880986939,
+      "grad_norm": 1.0018744468688965,
+      "learning_rate": 2.8145708817942606e-05,
+      "loss": 0.575,
+      "step": 350
+    },
+    {
+      "epoch": 1.0290275761973875,
+      "grad_norm": 0.949688196182251,
+      "learning_rate": 2.8075826517398523e-05,
+      "loss": 0.5709,
+      "step": 355
+    },
+    {
+      "epoch": 1.0435413642960814,
+      "grad_norm": 1.0343447923660278,
+      "learning_rate": 2.8004741737078305e-05,
+      "loss": 0.6286,
+      "step": 360
+    },
+    {
+      "epoch": 1.058055152394775,
+      "grad_norm": 1.0741147994995117,
+      "learning_rate": 2.7932461014083875e-05,
+      "loss": 0.5388,
+      "step": 365
+    },
+    {
+      "epoch": 1.0725689404934688,
+      "grad_norm": 1.1340123414993286,
+      "learning_rate": 2.7858990995498505e-05,
+      "loss": 0.6193,
+      "step": 370
+    },
+    {
+      "epoch": 1.0870827285921625,
+      "grad_norm": 1.3583769798278809,
+      "learning_rate": 2.7784338437775502e-05,
+      "loss": 0.5577,
+      "step": 375
+    },
+    {
+      "epoch": 1.1015965166908563,
+      "grad_norm": 1.2563444375991821,
+      "learning_rate": 2.770851020611692e-05,
+      "loss": 0.6166,
+      "step": 380
+    },
+    {
+      "epoch": 1.11611030478955,
+      "grad_norm": 1.0581562519073486,
+      "learning_rate": 2.7631513273842178e-05,
+      "loss": 0.5249,
+      "step": 385
+    },
+    {
+      "epoch": 1.1306240928882438,
+      "grad_norm": 1.0436842441558838,
+      "learning_rate": 2.7553354721746797e-05,
+      "loss": 0.5518,
+      "step": 390
+    },
+    {
+      "epoch": 1.1451378809869377,
+      "grad_norm": 1.1011406183242798,
+      "learning_rate": 2.747404173745124e-05,
+      "loss": 0.5251,
+      "step": 395
+    },
+    {
+      "epoch": 1.1596516690856313,
+      "grad_norm": 0.9995150566101074,
+      "learning_rate": 2.7393581614739924e-05,
+      "loss": 0.551,
+      "step": 400
+    },
+    {
+      "epoch": 1.1741654571843252,
+      "grad_norm": 1.0945287942886353,
+      "learning_rate": 2.7311981752890446e-05,
+      "loss": 0.5266,
+      "step": 405
+    },
+    {
+      "epoch": 1.1886792452830188,
+      "grad_norm": 1.1231412887573242,
+      "learning_rate": 2.7229249655993164e-05,
+      "loss": 0.5995,
+      "step": 410
+    },
+    {
+      "epoch": 1.2031930333817127,
+      "grad_norm": 1.0969406366348267,
+      "learning_rate": 2.7145392932261085e-05,
+      "loss": 0.5056,
+      "step": 415
+    },
+    {
+      "epoch": 1.2177068214804063,
+      "grad_norm": 1.0740689039230347,
+      "learning_rate": 2.706041929333019e-05,
+      "loss": 0.5104,
+      "step": 420
+    },
+    {
+      "epoch": 1.2322206095791002,
+      "grad_norm": 1.0396003723144531,
+      "learning_rate": 2.697433655355029e-05,
+      "loss": 0.5284,
+      "step": 425
+    },
+    {
+      "epoch": 1.2467343976777938,
+      "grad_norm": 0.9959568381309509,
+      "learning_rate": 2.6887152629266354e-05,
+      "loss": 0.518,
+      "step": 430
+    },
+    {
+      "epoch": 1.2612481857764877,
+      "grad_norm": 1.1106600761413574,
+      "learning_rate": 2.6798875538090565e-05,
+      "loss": 0.4998,
+      "step": 435
+    },
+    {
+      "epoch": 1.2757619738751815,
+      "grad_norm": 1.093956708908081,
+      "learning_rate": 2.6709513398164938e-05,
+      "loss": 0.5424,
+      "step": 440
+    },
+    {
+      "epoch": 1.2902757619738752,
+      "grad_norm": 1.2593542337417603,
+      "learning_rate": 2.6619074427414817e-05,
+      "loss": 0.5224,
+      "step": 445
+    },
+    {
+      "epoch": 1.3047895500725688,
+      "grad_norm": 1.1243525743484497,
+      "learning_rate": 2.6527566942793105e-05,
+      "loss": 0.5213,
+      "step": 450
+    },
+    {
+      "epoch": 1.3193033381712627,
+      "grad_norm": 1.195202112197876,
+      "learning_rate": 2.643499935951542e-05,
+      "loss": 0.5096,
+      "step": 455
+    },
+    {
+      "epoch": 1.3338171262699565,
+      "grad_norm": 1.229646921157837,
+      "learning_rate": 2.6341380190286233e-05,
+      "loss": 0.5101,
+      "step": 460
+    },
+    {
+      "epoch": 1.3483309143686502,
+      "grad_norm": 1.1680142879486084,
+      "learning_rate": 2.6246718044516013e-05,
+      "loss": 0.444,
+      "step": 465
+    },
+    {
+      "epoch": 1.362844702467344,
+      "grad_norm": 1.0221312046051025,
+      "learning_rate": 2.615102162752948e-05,
+      "loss": 0.5211,
+      "step": 470
+    },
+    {
+      "epoch": 1.3773584905660377,
+      "grad_norm": 1.2264982461929321,
+      "learning_rate": 2.605429973976505e-05,
+      "loss": 0.4746,
+      "step": 475
+    },
+    {
+      "epoch": 1.3918722786647315,
+      "grad_norm": 1.2884243726730347,
+      "learning_rate": 2.5956561275965548e-05,
+      "loss": 0.4915,
+      "step": 480
+    },
+    {
+      "epoch": 1.4063860667634254,
+      "grad_norm": 1.1846914291381836,
+      "learning_rate": 2.585781522436019e-05,
+      "loss": 0.468,
+      "step": 485
+    },
+    {
+      "epoch": 1.420899854862119,
+      "grad_norm": 1.09394371509552,
+      "learning_rate": 2.575807066583805e-05,
+      "loss": 0.4837,
+      "step": 490
+    },
+    {
+      "epoch": 1.4354136429608126,
+      "grad_norm": 1.1531438827514648,
+      "learning_rate": 2.565733677311292e-05,
+      "loss": 0.4818,
+      "step": 495
+    },
+    {
+      "epoch": 1.4499274310595065,
+      "grad_norm": 1.1869271993637085,
+      "learning_rate": 2.555562280987982e-05,
+      "loss": 0.4815,
+      "step": 500
+    },
+    {
+      "epoch": 1.4644412191582004,
+      "grad_norm": 0.9555249810218811,
+      "learning_rate": 2.545293812996303e-05,
+      "loss": 0.4516,
+      "step": 505
+    },
+    {
+      "epoch": 1.478955007256894,
+      "grad_norm": 1.2877460718154907,
+      "learning_rate": 2.534929217645594e-05,
+      "loss": 0.4291,
+      "step": 510
+    },
+    {
+      "epoch": 1.4934687953555879,
+      "grad_norm": 1.2242249250411987,
+      "learning_rate": 2.5244694480852624e-05,
+      "loss": 0.4653,
+      "step": 515
+    },
+    {
+      "epoch": 1.5079825834542815,
+      "grad_norm": 1.1914358139038086,
+      "learning_rate": 2.5139154662171316e-05,
+      "loss": 0.4862,
+      "step": 520
+    },
+    {
+      "epoch": 1.5224963715529753,
+      "grad_norm": 1.0356124639511108,
+      "learning_rate": 2.5032682426069803e-05,
+      "loss": 0.4527,
+      "step": 525
+    },
+    {
+      "epoch": 1.5370101596516692,
+      "grad_norm": 1.0836732387542725,
+      "learning_rate": 2.4925287563952894e-05,
+      "loss": 0.4595,
+      "step": 530
+    },
+    {
+      "epoch": 1.5515239477503628,
+      "grad_norm": 1.0991663932800293,
+      "learning_rate": 2.4816979952071974e-05,
+      "loss": 0.4244,
+      "step": 535
+    },
+    {
+      "epoch": 1.5660377358490565,
+      "grad_norm": 1.0540759563446045,
+      "learning_rate": 2.4707769550616762e-05,
+      "loss": 0.4281,
+      "step": 540
+    },
+    {
+      "epoch": 1.5805515239477503,
+      "grad_norm": 1.142854928970337,
+      "learning_rate": 2.4597666402799363e-05,
+      "loss": 0.4506,
+      "step": 545
+    },
+    {
+      "epoch": 1.5950653120464442,
+      "grad_norm": 1.1945523023605347,
+      "learning_rate": 2.4486680633930658e-05,
+      "loss": 0.4435,
+      "step": 550
+    },
+    {
+      "epoch": 1.6095791001451378,
+      "grad_norm": 1.200697898864746,
+      "learning_rate": 2.4374822450489173e-05,
+      "loss": 0.4331,
+      "step": 555
+    },
+    {
+      "epoch": 1.6240928882438317,
+      "grad_norm": 1.0646142959594727,
+      "learning_rate": 2.4262102139182453e-05,
+      "loss": 0.4211,
+      "step": 560
+    },
+    {
+      "epoch": 1.6386066763425253,
+      "grad_norm": 1.0593570470809937,
+      "learning_rate": 2.414853006600109e-05,
+      "loss": 0.397,
+      "step": 565
+    },
+    {
+      "epoch": 1.6531204644412192,
+      "grad_norm": 1.0598646402359009,
+      "learning_rate": 2.403411667526546e-05,
+      "loss": 0.408,
+      "step": 570
+    },
+    {
+      "epoch": 1.667634252539913,
+      "grad_norm": 1.0258424282073975,
+      "learning_rate": 2.3918872488665195e-05,
+      "loss": 0.374,
+      "step": 575
+    },
+    {
+      "epoch": 1.6821480406386067,
+      "grad_norm": 1.2203288078308105,
+      "learning_rate": 2.380280810429164e-05,
+      "loss": 0.4155,
+      "step": 580
+    },
+    {
+      "epoch": 1.6966618287373003,
+      "grad_norm": 1.2701225280761719,
+      "learning_rate": 2.3685934195663184e-05,
+      "loss": 0.3916,
+      "step": 585
+    },
+    {
+      "epoch": 1.7111756168359942,
+      "grad_norm": 1.2071703672409058,
+      "learning_rate": 2.3568261510743743e-05,
+      "loss": 0.3882,
+      "step": 590
+    },
+    {
+      "epoch": 1.725689404934688,
+      "grad_norm": 1.0981299877166748,
+      "learning_rate": 2.344980087095433e-05,
+      "loss": 0.4114,
+      "step": 595
+    },
+    {
+      "epoch": 1.7402031930333817,
+      "grad_norm": 1.2279996871948242,
+      "learning_rate": 2.3330563170177896e-05,
+      "loss": 0.4341,
+      "step": 600
+    },
+    {
+      "epoch": 1.7547169811320755,
+      "grad_norm": 1.215522050857544,
+      "learning_rate": 2.3210559373757515e-05,
+      "loss": 0.4438,
+      "step": 605
+    },
+    {
+      "epoch": 1.7692307692307692,
+      "grad_norm": 1.167236566543579,
+      "learning_rate": 2.308980051748799e-05,
+      "loss": 0.3922,
+      "step": 610
+    },
+    {
+      "epoch": 1.783744557329463,
+      "grad_norm": 1.12175714969635,
+      "learning_rate": 2.296829770660096e-05,
+      "loss": 0.3864,
+      "step": 615
+    },
+    {
+      "epoch": 1.7982583454281569,
+      "grad_norm": 1.1657911539077759,
+      "learning_rate": 2.2846062114743665e-05,
+      "loss": 0.3834,
+      "step": 620
+    },
+    {
+      "epoch": 1.8127721335268505,
+      "grad_norm": 1.0777708292007446,
+      "learning_rate": 2.272310498295138e-05,
+      "loss": 0.388,
+      "step": 625
+    },
+    {
+      "epoch": 1.8272859216255442,
+      "grad_norm": 1.0510711669921875,
+      "learning_rate": 2.259943761861367e-05,
+      "loss": 0.3934,
+      "step": 630
+    },
+    {
+      "epoch": 1.841799709724238,
+      "grad_norm": 1.141183853149414,
+      "learning_rate": 2.247507139443454e-05,
+      "loss": 0.3804,
+      "step": 635
+    },
+    {
+      "epoch": 1.8563134978229319,
+      "grad_norm": 1.1873807907104492,
+      "learning_rate": 2.2350017747386573e-05,
+      "loss": 0.3764,
+      "step": 640
+    },
+    {
+      "epoch": 1.8708272859216255,
+      "grad_norm": 1.1094430685043335,
+      "learning_rate": 2.2224288177659174e-05,
+      "loss": 0.3557,
+      "step": 645
+    },
+    {
+      "epoch": 1.8853410740203191,
+      "grad_norm": 1.0801291465759277,
+      "learning_rate": 2.2097894247600965e-05,
+      "loss": 0.333,
+      "step": 650
+    },
+    {
+      "epoch": 1.899854862119013,
+      "grad_norm": 1.171615719795227,
+      "learning_rate": 2.197084758065653e-05,
+      "loss": 0.3844,
+      "step": 655
+    },
+    {
+      "epoch": 1.9143686502177069,
+      "grad_norm": 1.1228846311569214,
+      "learning_rate": 2.1843159860297445e-05,
+      "loss": 0.3729,
+      "step": 660
+    },
+    {
+      "epoch": 1.9288824383164007,
+      "grad_norm": 1.0446889400482178,
+      "learning_rate": 2.171484282894789e-05,
+      "loss": 0.3764,
+      "step": 665
+    },
+    {
+      "epoch": 1.9433962264150944,
+      "grad_norm": 1.055737853050232,
+      "learning_rate": 2.158590828690477e-05,
+      "loss": 0.3313,
+      "step": 670
+    },
+    {
+      "epoch": 1.957910014513788,
+      "grad_norm": 1.2090764045715332,
+      "learning_rate": 2.1456368091252526e-05,
+      "loss": 0.3725,
+      "step": 675
+    },
+    {
+      "epoch": 1.9724238026124818,
+      "grad_norm": 1.0745823383331299,
+      "learning_rate": 2.1326234154772752e-05,
+      "loss": 0.3492,
+      "step": 680
+    },
+    {
+      "epoch": 1.9869375907111757,
+      "grad_norm": 1.2478907108306885,
+      "learning_rate": 2.1195518444848673e-05,
+      "loss": 0.3344,
+      "step": 685
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.450290560722351,
+      "learning_rate": 2.1064232982364572e-05,
+      "loss": 0.3748,
+      "step": 690
+    },
+    {
+      "epoch": 2.014513788098694,
+      "grad_norm": 1.1875301599502563,
+      "learning_rate": 2.093238984060037e-05,
+      "loss": 0.2611,
+      "step": 695
+    },
+    {
+      "epoch": 2.0290275761973877,
+      "grad_norm": 1.0669788122177124,
+      "learning_rate": 2.0800001144121284e-05,
+      "loss": 0.267,
+      "step": 700
+    },
+    {
+      "epoch": 2.043541364296081,
+      "grad_norm": 1.1223045587539673,
+      "learning_rate": 2.0667079067662882e-05,
+      "loss": 0.2914,
+      "step": 705
+    },
+    {
+      "epoch": 2.058055152394775,
+      "grad_norm": 0.9819235801696777,
+      "learning_rate": 2.0533635835011448e-05,
+      "loss": 0.3139,
+      "step": 710
+    },
+    {
+      "epoch": 2.072568940493469,
+      "grad_norm": 1.3752785921096802,
+      "learning_rate": 2.0399683717879837e-05,
+      "loss": 0.2765,
+      "step": 715
+    },
+    {
+      "epoch": 2.0870827285921627,
+      "grad_norm": 1.0086604356765747,
+      "learning_rate": 2.0265235034778983e-05,
+      "loss": 0.3018,
+      "step": 720
+    },
+    {
+      "epoch": 2.101596516690856,
+      "grad_norm": 1.1323165893554688,
+      "learning_rate": 2.0130302149885033e-05,
+      "loss": 0.3085,
+      "step": 725
+    },
+    {
+      "epoch": 2.11611030478955,
+      "grad_norm": 0.9608115553855896,
+      "learning_rate": 1.9994897471902325e-05,
+      "loss": 0.2563,
+      "step": 730
+    },
+    {
+      "epoch": 2.130624092888244,
+      "grad_norm": 1.293845534324646,
+      "learning_rate": 1.9859033452922253e-05,
+      "loss": 0.2519,
+      "step": 735
+    },
+    {
+      "epoch": 2.1451378809869377,
+      "grad_norm": 1.0907076597213745,
+      "learning_rate": 1.972272258727815e-05,
+      "loss": 0.2778,
+      "step": 740
+    },
+    {
+      "epoch": 2.1596516690856316,
+      "grad_norm": 1.265874981880188,
+      "learning_rate": 1.9585977410396295e-05,
+      "loss": 0.2577,
+      "step": 745
+    },
+    {
+      "epoch": 2.174165457184325,
+      "grad_norm": 1.1681596040725708,
+      "learning_rate": 1.9448810497643095e-05,
+      "loss": 0.2465,
+      "step": 750
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 1.2213428020477295,
+      "learning_rate": 1.9311234463168666e-05,
+      "loss": 0.259,
+      "step": 755
+    },
+    {
+      "epoch": 2.2031930333817127,
+      "grad_norm": 1.3326306343078613,
+      "learning_rate": 1.9173261958746793e-05,
+      "loss": 0.248,
+      "step": 760
+    },
+    {
+      "epoch": 2.2177068214804065,
+      "grad_norm": 1.1038895845413208,
+      "learning_rate": 1.9034905672611445e-05,
+      "loss": 0.2512,
+      "step": 765
+    },
+    {
+      "epoch": 2.2322206095791,
+      "grad_norm": 1.180465817451477,
+      "learning_rate": 1.889617832828995e-05,
+      "loss": 0.2609,
+      "step": 770
+    },
+    {
+      "epoch": 2.246734397677794,
+      "grad_norm": 1.1591984033584595,
+      "learning_rate": 1.8757092683432903e-05,
+      "loss": 0.2935,
+      "step": 775
+    },
+    {
+      "epoch": 2.2612481857764877,
+      "grad_norm": 1.2751474380493164,
+      "learning_rate": 1.8617661528640946e-05,
+      "loss": 0.2471,
+      "step": 780
+    },
+    {
+      "epoch": 2.2757619738751815,
+      "grad_norm": 1.1469013690948486,
+      "learning_rate": 1.8477897686288523e-05,
+      "loss": 0.2455,
+      "step": 785
+    },
+    {
+      "epoch": 2.2902757619738754,
+      "grad_norm": 1.1740764379501343,
+      "learning_rate": 1.8337814009344716e-05,
+      "loss": 0.2753,
+      "step": 790
+    },
+    {
+      "epoch": 2.304789550072569,
+      "grad_norm": 1.2813502550125122,
+      "learning_rate": 1.819742338019124e-05,
+      "loss": 0.2476,
+      "step": 795
+    },
+    {
+      "epoch": 2.3193033381712627,
+      "grad_norm": 1.2389832735061646,
+      "learning_rate": 1.8056738709437775e-05,
+      "loss": 0.2408,
+      "step": 800
+    },
+    {
+      "epoch": 2.3338171262699565,
+      "grad_norm": 1.0330638885498047,
+      "learning_rate": 1.7915772934734653e-05,
+      "loss": 0.2375,
+      "step": 805
+    },
+    {
+      "epoch": 2.3483309143686504,
+      "grad_norm": 1.3108922243118286,
+      "learning_rate": 1.7774539019583108e-05,
+      "loss": 0.2407,
+      "step": 810
+    },
+    {
+      "epoch": 2.362844702467344,
+      "grad_norm": 1.1805189847946167,
+      "learning_rate": 1.763304995214313e-05,
+      "loss": 0.2445,
+      "step": 815
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 1.2893935441970825,
+      "learning_rate": 1.7491318744039012e-05,
+      "loss": 0.2551,
+      "step": 820
+    },
+    {
+      "epoch": 2.3918722786647315,
+      "grad_norm": 1.0978128910064697,
+      "learning_rate": 1.734935842916282e-05,
+      "loss": 0.2377,
+      "step": 825
+    },
+    {
+      "epoch": 2.4063860667634254,
+      "grad_norm": 1.1712990999221802,
+      "learning_rate": 1.720718206247574e-05,
+      "loss": 0.2334,
+      "step": 830
+    },
+    {
+      "epoch": 2.4208998548621192,
+      "grad_norm": 1.3427225351333618,
+      "learning_rate": 1.7064802718807526e-05,
+      "loss": 0.2321,
+      "step": 835
+    },
+    {
+      "epoch": 2.4354136429608126,
+      "grad_norm": 1.1522012948989868,
+      "learning_rate": 1.69222334916541e-05,
+      "loss": 0.2443,
+      "step": 840
+    },
+    {
+      "epoch": 2.4499274310595065,
+      "grad_norm": 1.1092753410339355,
+      "learning_rate": 1.6779487491973472e-05,
+      "loss": 0.2327,
+      "step": 845
+    },
+    {
+      "epoch": 2.4644412191582004,
+      "grad_norm": 1.0141578912734985,
+      "learning_rate": 1.6636577846980014e-05,
+      "loss": 0.2435,
+      "step": 850
+    },
+    {
+      "epoch": 2.478955007256894,
+      "grad_norm": 1.3944146633148193,
+      "learning_rate": 1.6493517698937253e-05,
+      "loss": 0.2231,
+      "step": 855
+    },
+    {
+      "epoch": 2.4934687953555876,
+      "grad_norm": 1.0414365530014038,
+      "learning_rate": 1.6350320203949274e-05,
+      "loss": 0.2245,
+      "step": 860
+    },
+    {
+      "epoch": 2.5079825834542815,
+      "grad_norm": 1.1255793571472168,
+      "learning_rate": 1.6206998530750893e-05,
+      "loss": 0.2081,
+      "step": 865
+    },
+    {
+      "epoch": 2.5224963715529753,
+      "grad_norm": 1.2589868307113647,
+      "learning_rate": 1.6063565859496577e-05,
+      "loss": 0.232,
+      "step": 870
+    },
+    {
+      "epoch": 2.537010159651669,
+      "grad_norm": 1.2401530742645264,
+      "learning_rate": 1.5920035380548433e-05,
+      "loss": 0.216,
+      "step": 875
+    },
+    {
+      "epoch": 2.551523947750363,
+      "grad_norm": 1.092507243156433,
+      "learning_rate": 1.5776420293263144e-05,
+      "loss": 0.2228,
+      "step": 880
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 1.1913965940475464,
+      "learning_rate": 1.5632733804778154e-05,
+      "loss": 0.2006,
+      "step": 885
+    },
+    {
+      "epoch": 2.5805515239477503,
+      "grad_norm": 1.0700242519378662,
+      "learning_rate": 1.54889891287971e-05,
+      "loss": 0.2449,
+      "step": 890
+    },
+    {
+      "epoch": 2.595065312046444,
+      "grad_norm": 1.0306884050369263,
+      "learning_rate": 1.534519948437468e-05,
+      "loss": 0.1815,
+      "step": 895
+    },
+    {
+      "epoch": 2.6095791001451376,
+      "grad_norm": 1.108221173286438,
+      "learning_rate": 1.5201378094700963e-05,
+      "loss": 0.2089,
+      "step": 900
+    },
+    {
+      "epoch": 2.6240928882438315,
+      "grad_norm": 1.1428337097167969,
+      "learning_rate": 1.5057538185885385e-05,
+      "loss": 0.1899,
+      "step": 905
+    },
+    {
+      "epoch": 2.6386066763425253,
+      "grad_norm": 1.193100094795227,
+      "learning_rate": 1.4913692985740438e-05,
+      "loss": 0.2571,
+      "step": 910
+    },
+    {
+      "epoch": 2.653120464441219,
+      "grad_norm": 1.167102575302124,
+      "learning_rate": 1.4769855722565214e-05,
+      "loss": 0.25,
+      "step": 915
+    },
+    {
+      "epoch": 2.667634252539913,
+      "grad_norm": 1.152069091796875,
+      "learning_rate": 1.462603962392891e-05,
+      "loss": 0.2327,
+      "step": 920
+    },
+    {
+      "epoch": 2.682148040638607,
+      "grad_norm": 1.0608423948287964,
+      "learning_rate": 1.4482257915454383e-05,
+      "loss": 0.2055,
+      "step": 925
+    },
+    {
+      "epoch": 2.6966618287373003,
+      "grad_norm": 1.1504971981048584,
+      "learning_rate": 1.43385238196019e-05,
+      "loss": 0.2048,
+      "step": 930
+    },
+    {
+      "epoch": 2.711175616835994,
+      "grad_norm": 1.1371382474899292,
+      "learning_rate": 1.4194850554453175e-05,
+      "loss": 0.2271,
+      "step": 935
+    },
+    {
+      "epoch": 2.725689404934688,
+      "grad_norm": 1.0657782554626465,
+      "learning_rate": 1.4051251332495819e-05,
+      "loss": 0.2092,
+      "step": 940
+    },
+    {
+      "epoch": 2.7402031930333814,
+      "grad_norm": 1.3242257833480835,
+      "learning_rate": 1.3907739359408259e-05,
+      "loss": 0.2295,
+      "step": 945
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 1.1220176219940186,
+      "learning_rate": 1.3764327832845344e-05,
+      "loss": 0.1953,
+      "step": 950
+    },
+    {
+      "epoch": 2.769230769230769,
+      "grad_norm": 1.0856845378875732,
+      "learning_rate": 1.3621029941224666e-05,
+      "loss": 0.2045,
+      "step": 955
+    },
+    {
+      "epoch": 2.783744557329463,
+      "grad_norm": 1.239922046661377,
+      "learning_rate": 1.3477858862513698e-05,
+      "loss": 0.203,
+      "step": 960
+    },
+    {
+      "epoch": 2.798258345428157,
+      "grad_norm": 1.0360215902328491,
+      "learning_rate": 1.3334827763017952e-05,
+      "loss": 0.2098,
+      "step": 965
+    },
+    {
+      "epoch": 2.8127721335268507,
+      "grad_norm": 1.1175003051757812,
+      "learning_rate": 1.3191949796170156e-05,
+      "loss": 0.1812,
+      "step": 970
+    },
+    {
+      "epoch": 2.827285921625544,
+      "grad_norm": 1.1741539239883423,
+      "learning_rate": 1.3049238101320652e-05,
+      "loss": 0.2123,
+      "step": 975
+    },
+    {
+      "epoch": 2.841799709724238,
+      "grad_norm": 1.0360662937164307,
+      "learning_rate": 1.2906705802529064e-05,
+      "loss": 0.1724,
+      "step": 980
+    },
+    {
+      "epoch": 2.856313497822932,
+      "grad_norm": 1.0688581466674805,
+      "learning_rate": 1.2764366007357382e-05,
+      "loss": 0.174,
+      "step": 985
+    },
+    {
+      "epoch": 2.8708272859216253,
+      "grad_norm": 1.1229158639907837,
+      "learning_rate": 1.2622231805664592e-05,
+      "loss": 0.2262,
+      "step": 990
+    },
+    {
+      "epoch": 2.885341074020319,
+      "grad_norm": 1.1505873203277588,
+      "learning_rate": 1.2480316268402859e-05,
+      "loss": 0.1717,
+      "step": 995
+    },
+    {
+      "epoch": 2.899854862119013,
+      "grad_norm": 1.1841315031051636,
+      "learning_rate": 1.2338632446415526e-05,
+      "loss": 0.1993,
+      "step": 1000
+    },
+    {
+      "epoch": 2.914368650217707,
+      "grad_norm": 1.1179850101470947,
+      "learning_rate": 1.2197193369236929e-05,
+      "loss": 0.1783,
+      "step": 1005
+    },
+    {
+      "epoch": 2.9288824383164007,
+      "grad_norm": 1.2388747930526733,
+      "learning_rate": 1.2056012043894166e-05,
+      "loss": 0.1867,
+      "step": 1010
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 1.0603691339492798,
+      "learning_rate": 1.1915101453710953e-05,
+      "loss": 0.1724,
+      "step": 1015
+    },
+    {
+      "epoch": 2.957910014513788,
+      "grad_norm": 1.209848165512085,
+      "learning_rate": 1.1774474557113642e-05,
+      "loss": 0.2029,
+      "step": 1020
+    },
+    {
+      "epoch": 2.972423802612482,
+      "grad_norm": 1.1248085498809814,
+      "learning_rate": 1.163414428643954e-05,
+      "loss": 0.2085,
+      "step": 1025
+    },
+    {
+      "epoch": 2.9869375907111757,
+      "grad_norm": 1.0725529193878174,
+      "learning_rate": 1.1494123546747638e-05,
+      "loss": 0.1946,
+      "step": 1030
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.5596674680709839,
+      "learning_rate": 1.1354425214631804e-05,
+      "loss": 0.1845,
+      "step": 1035
+    },
+    {
+      "epoch": 3.014513788098694,
+      "grad_norm": 0.9024957418441772,
+      "learning_rate": 1.1215062137036673e-05,
+      "loss": 0.1371,
+      "step": 1040
+    },
+    {
+      "epoch": 3.0290275761973877,
+      "grad_norm": 1.1488008499145508,
+      "learning_rate": 1.1076047130076169e-05,
+      "loss": 0.1233,
+      "step": 1045
+    },
+    {
+      "epoch": 3.043541364296081,
+      "grad_norm": 0.9522742033004761,
+      "learning_rate": 1.0937392977854926e-05,
+      "loss": 0.1342,
+      "step": 1050
+    },
+    {
+      "epoch": 3.058055152394775,
+      "grad_norm": 0.913542628288269,
+      "learning_rate": 1.0799112431292642e-05,
+      "loss": 0.1574,
+      "step": 1055
+    },
+    {
+      "epoch": 3.072568940493469,
+      "grad_norm": 1.0045816898345947,
+      "learning_rate": 1.0661218206951486e-05,
+      "loss": 0.1402,
+      "step": 1060
+    },
+    {
+      "epoch": 3.0870827285921627,
+      "grad_norm": 0.8637295365333557,
+      "learning_rate": 1.0523722985866625e-05,
+      "loss": 0.1294,
+      "step": 1065
+    },
+    {
+      "epoch": 3.101596516690856,
+      "grad_norm": 1.6149895191192627,
+      "learning_rate": 1.038663941238008e-05,
+      "loss": 0.1776,
+      "step": 1070
+    },
+    {
+      "epoch": 3.11611030478955,
+      "grad_norm": 0.9293178915977478,
+      "learning_rate": 1.0249980092977916e-05,
+      "loss": 0.1384,
+      "step": 1075
+    },
+    {
+      "epoch": 3.130624092888244,
+      "grad_norm": 1.0604490041732788,
+      "learning_rate": 1.0113757595130917e-05,
+      "loss": 0.1392,
+      "step": 1080
+    },
+    {
+      "epoch": 3.1451378809869377,
+      "grad_norm": 1.013123869895935,
+      "learning_rate": 9.977984446138875e-06,
+      "loss": 0.149,
+      "step": 1085
+    },
+    {
+      "epoch": 3.1596516690856316,
+      "grad_norm": 1.0887017250061035,
+      "learning_rate": 9.842673131978522e-06,
+      "loss": 0.1155,
+      "step": 1090
+    },
+    {
+      "epoch": 3.174165457184325,
+      "grad_norm": 1.1860378980636597,
+      "learning_rate": 9.707836096155342e-06,
+      "loss": 0.127,
+      "step": 1095
+    },
+    {
+      "epoch": 3.188679245283019,
+      "grad_norm": 1.0214670896530151,
+      "learning_rate": 9.573485738559187e-06,
+      "loss": 0.1426,
+      "step": 1100
+    },
+    {
+      "epoch": 3.2031930333817127,
+      "grad_norm": 1.0829511880874634,
+      "learning_rate": 9.439634414323987e-06,
+      "loss": 0.1449,
+      "step": 1105
+    },
+    {
+      "epoch": 3.2177068214804065,
+      "grad_norm": 0.9733559489250183,
+      "learning_rate": 9.306294432691556e-06,
+      "loss": 0.1344,
+      "step": 1110
+    },
+    {
+      "epoch": 3.2322206095791,
+      "grad_norm": 0.9255797863006592,
+      "learning_rate": 9.17347805587958e-06,
+      "loss": 0.1184,
+      "step": 1115
+    },
+    {
+      "epoch": 3.246734397677794,
+      "grad_norm": 0.797142744064331,
+      "learning_rate": 9.041197497953984e-06,
+      "loss": 0.1483,
+      "step": 1120
+    },
+    {
+      "epoch": 3.2612481857764877,
+      "grad_norm": 1.1751822233200073,
+      "learning_rate": 8.909464923705685e-06,
+      "loss": 0.1452,
+      "step": 1125
+    },
+    {
+      "epoch": 3.2757619738751815,
+      "grad_norm": 0.9948896765708923,
+      "learning_rate": 8.7782924475319e-06,
+      "loss": 0.1229,
+      "step": 1130
+    },
+    {
+      "epoch": 3.2902757619738754,
+      "grad_norm": 1.0014498233795166,
+      "learning_rate": 8.647692132322084e-06,
+      "loss": 0.1297,
+      "step": 1135
+    },
+    {
+      "epoch": 3.304789550072569,
+      "grad_norm": 1.0062052011489868,
+      "learning_rate": 8.517675988348592e-06,
+      "loss": 0.1474,
+      "step": 1140
+    },
+    {
+      "epoch": 3.3193033381712627,
+      "grad_norm": 1.0303725004196167,
+      "learning_rate": 8.38825597216221e-06,
+      "loss": 0.1281,
+      "step": 1145
+    },
+    {
+      "epoch": 3.3338171262699565,
+      "grad_norm": 0.9333248734474182,
+      "learning_rate": 8.259443985492577e-06,
+      "loss": 0.1228,
+      "step": 1150
+    },
+    {
+      "epoch": 3.3483309143686504,
+      "grad_norm": 1.212723970413208,
+      "learning_rate": 8.131251874153701e-06,
+      "loss": 0.124,
+      "step": 1155
+    },
+    {
+      "epoch": 3.362844702467344,
+      "grad_norm": 1.2014771699905396,
+      "learning_rate": 8.003691426954578e-06,
+      "loss": 0.1488,
+      "step": 1160
+    },
+    {
+      "epoch": 3.3773584905660377,
+      "grad_norm": 0.9629170298576355,
+      "learning_rate": 7.87677437461509e-06,
+      "loss": 0.1248,
+      "step": 1165
+    },
+    {
+      "epoch": 3.3918722786647315,
+      "grad_norm": 1.1353247165679932,
+      "learning_rate": 7.750512388687191e-06,
+      "loss": 0.1335,
+      "step": 1170
+    },
+    {
+      "epoch": 3.4063860667634254,
+      "grad_norm": 0.9715927243232727,
+      "learning_rate": 7.624917080481608e-06,
+      "loss": 0.1174,
+      "step": 1175
+    },
+    {
+      "epoch": 3.4208998548621192,
+      "grad_norm": 1.1411174535751343,
+      "learning_rate": 7.500000000000004e-06,
+      "loss": 0.1358,
+      "step": 1180
+    },
+    {
+      "epoch": 3.4354136429608126,
+      "grad_norm": 0.8531781435012817,
+      "learning_rate": 7.375772634872838e-06,
+      "loss": 0.1348,
+      "step": 1185
+    },
+    {
+      "epoch": 3.4499274310595065,
+      "grad_norm": 1.1151381731033325,
+      "learning_rate": 7.252246409302928e-06,
+      "loss": 0.1177,
+      "step": 1190
+    },
+    {
+      "epoch": 3.4644412191582004,
+      "grad_norm": 0.9225158095359802,
+      "learning_rate": 7.12943268301488e-06,
+      "loss": 0.1221,
+      "step": 1195
+    },
+    {
+      "epoch": 3.478955007256894,
+      "grad_norm": 0.9018508195877075,
+      "learning_rate": 7.007342750210386e-06,
+      "loss": 0.118,
+      "step": 1200
+    },
+    {
+      "epoch": 3.4934687953555876,
+      "grad_norm": 0.9326978325843811,
+      "learning_rate": 6.885987838529621e-06,
+      "loss": 0.123,
+      "step": 1205
+    },
+    {
+      "epoch": 3.5079825834542815,
+      "grad_norm": 0.9079959392547607,
+      "learning_rate": 6.765379108018704e-06,
+      "loss": 0.124,
+      "step": 1210
+    },
+    {
+      "epoch": 3.5224963715529753,
+      "grad_norm": 1.0523256063461304,
+      "learning_rate": 6.645527650103413e-06,
+      "loss": 0.1295,
+      "step": 1215
+    },
+    {
+      "epoch": 3.537010159651669,
+      "grad_norm": 0.932341992855072,
+      "learning_rate": 6.526444486569182e-06,
+      "loss": 0.1116,
+      "step": 1220
+    },
+    {
+      "epoch": 3.551523947750363,
+      "grad_norm": 1.014511227607727,
+      "learning_rate": 6.408140568547524e-06,
+      "loss": 0.1066,
+      "step": 1225
+    },
+    {
+      "epoch": 3.5660377358490565,
+      "grad_norm": 1.1238880157470703,
+      "learning_rate": 6.290626775508936e-06,
+      "loss": 0.1298,
+      "step": 1230
+    },
+    {
+      "epoch": 3.5805515239477503,
+      "grad_norm": 0.9214984774589539,
+      "learning_rate": 6.1739139142624165e-06,
+      "loss": 0.126,
+      "step": 1235
+    },
+    {
+      "epoch": 3.595065312046444,
+      "grad_norm": 0.7832999229431152,
+      "learning_rate": 6.058012717961621e-06,
+      "loss": 0.1216,
+      "step": 1240
+    },
+    {
+      "epoch": 3.6095791001451376,
+      "grad_norm": 0.9536693692207336,
+      "learning_rate": 5.942933845117836e-06,
+      "loss": 0.1318,
+      "step": 1245
+    },
+    {
+      "epoch": 3.6240928882438315,
+      "grad_norm": 0.8524763584136963,
+      "learning_rate": 5.828687878619816e-06,
+      "loss": 0.1264,
+      "step": 1250
+    },
+    {
+      "epoch": 3.6386066763425253,
+      "grad_norm": 0.9581038951873779,
+      "learning_rate": 5.715285324760525e-06,
+      "loss": 0.1214,
+      "step": 1255
+    },
+    {
+      "epoch": 3.653120464441219,
+      "grad_norm": 0.8654883503913879,
+      "learning_rate": 5.60273661227098e-06,
+      "loss": 0.1388,
+      "step": 1260
+    },
+    {
+      "epoch": 3.667634252539913,
+      "grad_norm": 0.909365177154541,
+      "learning_rate": 5.4910520913611975e-06,
+      "loss": 0.1086,
+      "step": 1265
+    },
+    {
+      "epoch": 3.682148040638607,
+      "grad_norm": 1.2761754989624023,
+      "learning_rate": 5.38024203276837e-06,
+      "loss": 0.1323,
+      "step": 1270
+    },
+    {
+      "epoch": 3.6966618287373003,
+      "grad_norm": 0.9908791184425354,
+      "learning_rate": 5.2703166268123395e-06,
+      "loss": 0.0975,
+      "step": 1275
+    },
+    {
+      "epoch": 3.711175616835994,
+      "grad_norm": 0.861386239528656,
+      "learning_rate": 5.161285982458486e-06,
+      "loss": 0.1118,
+      "step": 1280
+    },
+    {
+      "epoch": 3.725689404934688,
+      "grad_norm": 0.973271369934082,
+      "learning_rate": 5.0531601263880755e-06,
+      "loss": 0.1078,
+      "step": 1285
+    },
+    {
+      "epoch": 3.7402031930333814,
+      "grad_norm": 0.9156631827354431,
+      "learning_rate": 4.945949002076194e-06,
+      "loss": 0.1227,
+      "step": 1290
+    },
+    {
+      "epoch": 3.7547169811320753,
+      "grad_norm": 0.9608476161956787,
+      "learning_rate": 4.839662468877321e-06,
+      "loss": 0.1139,
+      "step": 1295
+    },
+    {
+      "epoch": 3.769230769230769,
+      "grad_norm": 0.8815176486968994,
+      "learning_rate": 4.734310301118652e-06,
+      "loss": 0.1075,
+      "step": 1300
+    },
+    {
+      "epoch": 3.783744557329463,
+      "grad_norm": 0.9263197779655457,
+      "learning_rate": 4.629902187201213e-06,
+      "loss": 0.1034,
+      "step": 1305
+    },
+    {
+      "epoch": 3.798258345428157,
+      "grad_norm": 0.942035436630249,
+      "learning_rate": 4.526447728708909e-06,
+      "loss": 0.1134,
+      "step": 1310
+    },
+    {
+      "epoch": 3.8127721335268507,
+      "grad_norm": 0.8410844802856445,
+      "learning_rate": 4.423956439525544e-06,
+      "loss": 0.1024,
+      "step": 1315
+    },
+    {
+      "epoch": 3.827285921625544,
+      "grad_norm": 1.085524559020996,
+      "learning_rate": 4.322437744959894e-06,
+      "loss": 0.127,
+      "step": 1320
+    },
+    {
+      "epoch": 3.841799709724238,
+      "grad_norm": 0.882469892501831,
+      "learning_rate": 4.221900980878949e-06,
+      "loss": 0.0912,
+      "step": 1325
+    },
+    {
+      "epoch": 3.856313497822932,
+      "grad_norm": 0.7348304986953735,
+      "learning_rate": 4.122355392849357e-06,
+      "loss": 0.1143,
+      "step": 1330
+    },
+    {
+      "epoch": 3.8708272859216253,
+      "grad_norm": 0.8587133884429932,
+      "learning_rate": 4.023810135287185e-06,
+      "loss": 0.1023,
+      "step": 1335
+    },
+    {
+      "epoch": 3.885341074020319,
+      "grad_norm": 0.8341846466064453,
+      "learning_rate": 3.926274270616076e-06,
+      "loss": 0.1103,
+      "step": 1340
+    },
+    {
+      "epoch": 3.899854862119013,
+      "grad_norm": 0.962035059928894,
+      "learning_rate": 3.82975676843382e-06,
+      "loss": 0.1112,
+      "step": 1345
+    },
+    {
+      "epoch": 3.914368650217707,
+      "grad_norm": 0.9774994850158691,
+      "learning_rate": 3.7342665046875087e-06,
+      "loss": 0.1054,
+      "step": 1350
+    },
+    {
+      "epoch": 3.9288824383164007,
+      "grad_norm": 0.9391301870346069,
+      "learning_rate": 3.639812260857302e-06,
+      "loss": 0.117,
+      "step": 1355
+    },
+    {
+      "epoch": 3.9433962264150946,
+      "grad_norm": 0.8560742139816284,
+      "learning_rate": 3.5464027231488265e-06,
+      "loss": 0.1176,
+      "step": 1360
+    },
+    {
+      "epoch": 3.957910014513788,
+      "grad_norm": 0.7581481337547302,
+      "learning_rate": 3.4540464816944063e-06,
+      "loss": 0.1069,
+      "step": 1365
+    },
+    {
+      "epoch": 3.972423802612482,
+      "grad_norm": 0.9039914608001709,
+      "learning_rate": 3.3627520297630803e-06,
+      "loss": 0.1151,
+      "step": 1370
+    },
+    {
+      "epoch": 3.9869375907111757,
+      "grad_norm": 1.0366986989974976,
+      "learning_rate": 3.272527762979553e-06,
+      "loss": 0.1046,
+      "step": 1375
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.8571218848228455,
+      "learning_rate": 3.1833819785521113e-06,
+      "loss": 0.108,
+      "step": 1380
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1725,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.026963821616169e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

40_128_e5_3e-5/checkpoint-1380/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb03dfd1a09e76ed2bbbbab487b8a4cf3a040b6e2f6cfe9b561e1a7b4b2a3466
+size 7736

40_128_e5_3e-5/checkpoint-1380/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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 param_shapes.items():
+
+        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_file, 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

40_128_e5_3e-5/checkpoint-1725/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+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

40_128_e5_3e-5/checkpoint-1725/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

40_128_e5_3e-5/checkpoint-1725/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d37d97a0baa8e20a5f6ac9b327b7a6142d9747bacd179d5363cc53be7407f651
+size 791751704

40_128_e5_3e-5/checkpoint-1725/latest

@@ -0,0 +1 @@
+global_step1725

40_128_e5_3e-5/checkpoint-1725/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4516984cd034f887ee8311f2ca234eabd81eee0995e88836097a6517d3331710
+size 15920

40_128_e5_3e-5/checkpoint-1725/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa004fdabb75b02e0879e494b5c2117d5b76c879b683d81ec41b29849dce948f
+size 15920

40_128_e5_3e-5/checkpoint-1725/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1a6f2e24f4e22cd0ca81f99e8dd20f9fac64c86bb2c814c3f2b7a696d6a65e6
+size 15920