Upload folder using huggingface_hub

50_128_e5_3e-5/checkpoint-1000/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

50_128_e5_3e-5/checkpoint-1000/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": [
+    "o_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

50_128_e5_3e-5/checkpoint-1000/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ea73e855fac875c87f8afd3663b637c53b950b5eae0e41733d05dfab4cd9714
+size 791751704

50_128_e5_3e-5/checkpoint-1000/latest

@@ -0,0 +1 @@
+global_step1000

50_128_e5_3e-5/checkpoint-1000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91ad82de654a1235a3c6e4aef4218d6b17821b93f21be5e3fd0491fde00324cf
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bd9e0a773b4d0975b70b1aeab88f5a2f935dd82756f214d4e9c718c417291b0
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:575fd5206001e8e6d6c91343fd1f4f8572c03e51bb6a6876bdafb59df0fae654
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c81b33bffaae6a9795cafffc219d2605116667edd60cb60844b8c0688d59d280
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5824192066877aee10bd5916c780b223cc6942cc44d016809a59dd5560100798
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b11f5d5f45443d3b55966447f50ae3e6ebadfd3f86c71bdd0cad107175c5e23
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80ec871b8fb0c06495d379f6956020dfe9ba3bae4aa4208f3591dff08e91a73b
+size 15920

50_128_e5_3e-5/checkpoint-1000/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:586d6d90b4864e4c6bf334a30820f51273dce5ce0c26a5eaccf2818f40faa440
+size 15920

50_128_e5_3e-5/checkpoint-1000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:726d04b64afe038d27b90b442151c8ea243a733bc48f3067f65185d6315a0961
+size 1064

50_128_e5_3e-5/checkpoint-1000/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "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": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

50_128_e5_3e-5/checkpoint-1000/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": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

50_128_e5_3e-5/checkpoint-1000/trainer_state.json

@@ -0,0 +1,1434 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02004008016032064,
+      "grad_norm": 0.9890378713607788,
+      "learning_rate": 1.9047619047619047e-06,
+      "loss": 1.4301,
+      "step": 5
+    },
+    {
+      "epoch": 0.04008016032064128,
+      "grad_norm": 0.863508939743042,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 1.4641,
+      "step": 10
+    },
+    {
+      "epoch": 0.06012024048096192,
+      "grad_norm": 0.6787753701210022,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.4197,
+      "step": 15
+    },
+    {
+      "epoch": 0.08016032064128256,
+      "grad_norm": 0.6681222319602966,
+      "learning_rate": 9.047619047619047e-06,
+      "loss": 1.4312,
+      "step": 20
+    },
+    {
+      "epoch": 0.10020040080160321,
+      "grad_norm": 0.5663285255432129,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 1.3627,
+      "step": 25
+    },
+    {
+      "epoch": 0.12024048096192384,
+      "grad_norm": 0.5837746858596802,
+      "learning_rate": 1.380952380952381e-05,
+      "loss": 1.3997,
+      "step": 30
+    },
+    {
+      "epoch": 0.1402805611222445,
+      "grad_norm": 0.5583487153053284,
+      "learning_rate": 1.619047619047619e-05,
+      "loss": 1.3404,
+      "step": 35
+    },
+    {
+      "epoch": 0.16032064128256512,
+      "grad_norm": 0.4913857877254486,
+      "learning_rate": 1.8571428571428572e-05,
+      "loss": 1.3386,
+      "step": 40
+    },
+    {
+      "epoch": 0.18036072144288579,
+      "grad_norm": 0.573939323425293,
+      "learning_rate": 2.095238095238095e-05,
+      "loss": 1.3201,
+      "step": 45
+    },
+    {
+      "epoch": 0.20040080160320642,
+      "grad_norm": 0.508334219455719,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 1.2503,
+      "step": 50
+    },
+    {
+      "epoch": 0.22044088176352705,
+      "grad_norm": 0.5239691734313965,
+      "learning_rate": 2.5714285714285714e-05,
+      "loss": 1.2509,
+      "step": 55
+    },
+    {
+      "epoch": 0.24048096192384769,
+      "grad_norm": 0.7545399069786072,
+      "learning_rate": 2.8095238095238096e-05,
+      "loss": 1.2877,
+      "step": 60
+    },
+    {
+      "epoch": 0.2605210420841683,
+      "grad_norm": 0.4638131260871887,
+      "learning_rate": 2.9999947463720068e-05,
+      "loss": 1.219,
+      "step": 65
+    },
+    {
+      "epoch": 0.280561122244489,
+      "grad_norm": 0.5463809370994568,
+      "learning_rate": 2.9998108732563026e-05,
+      "loss": 1.2131,
+      "step": 70
+    },
+    {
+      "epoch": 0.30060120240480964,
+      "grad_norm": 0.5909023284912109,
+      "learning_rate": 2.9993643555405922e-05,
+      "loss": 1.1617,
+      "step": 75
+    },
+    {
+      "epoch": 0.32064128256513025,
+      "grad_norm": 0.4846280515193939,
+      "learning_rate": 2.9986552714183782e-05,
+      "loss": 1.2422,
+      "step": 80
+    },
+    {
+      "epoch": 0.3406813627254509,
+      "grad_norm": 0.6445205807685852,
+      "learning_rate": 2.9976837450633944e-05,
+      "loss": 1.1748,
+      "step": 85
+    },
+    {
+      "epoch": 0.36072144288577157,
+      "grad_norm": 0.6564697623252869,
+      "learning_rate": 2.996449946607859e-05,
+      "loss": 1.1738,
+      "step": 90
+    },
+    {
+      "epoch": 0.3807615230460922,
+      "grad_norm": 0.7153516411781311,
+      "learning_rate": 2.9949540921126824e-05,
+      "loss": 1.1438,
+      "step": 95
+    },
+    {
+      "epoch": 0.40080160320641284,
+      "grad_norm": 0.6607214212417603,
+      "learning_rate": 2.9931964435296292e-05,
+      "loss": 1.1229,
+      "step": 100
+    },
+    {
+      "epoch": 0.42084168336673344,
+      "grad_norm": 0.6068851351737976,
+      "learning_rate": 2.991177308655447e-05,
+      "loss": 1.1227,
+      "step": 105
+    },
+    {
+      "epoch": 0.4408817635270541,
+      "grad_norm": 0.6417985558509827,
+      "learning_rate": 2.988897041077966e-05,
+      "loss": 1.1068,
+      "step": 110
+    },
+    {
+      "epoch": 0.46092184368737477,
+      "grad_norm": 0.6902046203613281,
+      "learning_rate": 2.9863560401141773e-05,
+      "loss": 1.1214,
+      "step": 115
+    },
+    {
+      "epoch": 0.48096192384769537,
+      "grad_norm": 0.6481944918632507,
+      "learning_rate": 2.9835547507403067e-05,
+      "loss": 1.0765,
+      "step": 120
+    },
+    {
+      "epoch": 0.501002004008016,
+      "grad_norm": 0.7215368747711182,
+      "learning_rate": 2.980493663513891e-05,
+      "loss": 1.0368,
+      "step": 125
+    },
+    {
+      "epoch": 0.5210420841683366,
+      "grad_norm": 0.7066978216171265,
+      "learning_rate": 2.9771733144878706e-05,
+      "loss": 0.9668,
+      "step": 130
+    },
+    {
+      "epoch": 0.5410821643286573,
+      "grad_norm": 0.7489072680473328,
+      "learning_rate": 2.9735942851167202e-05,
+      "loss": 1.0239,
+      "step": 135
+    },
+    {
+      "epoch": 0.561122244488978,
+      "grad_norm": 0.7932925224304199,
+      "learning_rate": 2.9697572021546216e-05,
+      "loss": 0.9795,
+      "step": 140
+    },
+    {
+      "epoch": 0.5811623246492986,
+      "grad_norm": 0.9287444949150085,
+      "learning_rate": 2.9656627375457102e-05,
+      "loss": 0.9915,
+      "step": 145
+    },
+    {
+      "epoch": 0.6012024048096193,
+      "grad_norm": 0.8538782596588135,
+      "learning_rate": 2.961311608306403e-05,
+      "loss": 1.0117,
+      "step": 150
+    },
+    {
+      "epoch": 0.6212424849699398,
+      "grad_norm": 0.773285984992981,
+      "learning_rate": 2.956704576399838e-05,
+      "loss": 0.9697,
+      "step": 155
+    },
+    {
+      "epoch": 0.6412825651302605,
+      "grad_norm": 0.9210479855537415,
+      "learning_rate": 2.9518424486024382e-05,
+      "loss": 0.9589,
+      "step": 160
+    },
+    {
+      "epoch": 0.6613226452905812,
+      "grad_norm": 0.7253521680831909,
+      "learning_rate": 2.9467260763626323e-05,
+      "loss": 0.9559,
+      "step": 165
+    },
+    {
+      "epoch": 0.6813627254509018,
+      "grad_norm": 1.1397440433502197,
+      "learning_rate": 2.9413563556517483e-05,
+      "loss": 0.9394,
+      "step": 170
+    },
+    {
+      "epoch": 0.7014028056112225,
+      "grad_norm": 0.9524412751197815,
+      "learning_rate": 2.935734226807114e-05,
+      "loss": 0.9114,
+      "step": 175
+    },
+    {
+      "epoch": 0.7214428857715431,
+      "grad_norm": 0.8871064186096191,
+      "learning_rate": 2.9298606743673854e-05,
+      "loss": 0.865,
+      "step": 180
+    },
+    {
+      "epoch": 0.7414829659318637,
+      "grad_norm": 0.9255661368370056,
+      "learning_rate": 2.9237367269001362e-05,
+      "loss": 0.9108,
+      "step": 185
+    },
+    {
+      "epoch": 0.7615230460921844,
+      "grad_norm": 0.9735074639320374,
+      "learning_rate": 2.9173634568217366e-05,
+      "loss": 0.8465,
+      "step": 190
+    },
+    {
+      "epoch": 0.781563126252505,
+      "grad_norm": 0.9207878112792969,
+      "learning_rate": 2.9107419802095546e-05,
+      "loss": 0.8566,
+      "step": 195
+    },
+    {
+      "epoch": 0.8016032064128257,
+      "grad_norm": 0.9188950657844543,
+      "learning_rate": 2.9038734566065068e-05,
+      "loss": 0.815,
+      "step": 200
+    },
+    {
+      "epoch": 0.8216432865731463,
+      "grad_norm": 1.1056013107299805,
+      "learning_rate": 2.8967590888180052e-05,
+      "loss": 0.8299,
+      "step": 205
+    },
+    {
+      "epoch": 0.8416833667334669,
+      "grad_norm": 1.0484192371368408,
+      "learning_rate": 2.8894001227013213e-05,
+      "loss": 0.8266,
+      "step": 210
+    },
+    {
+      "epoch": 0.8617234468937875,
+      "grad_norm": 0.9629917144775391,
+      "learning_rate": 2.8817978469474137e-05,
+      "loss": 0.8386,
+      "step": 215
+    },
+    {
+      "epoch": 0.8817635270541082,
+      "grad_norm": 0.9744315147399902,
+      "learning_rate": 2.873953592855255e-05,
+      "loss": 0.7977,
+      "step": 220
+    },
+    {
+      "epoch": 0.9018036072144289,
+      "grad_norm": 1.0901851654052734,
+      "learning_rate": 2.8658687340986962e-05,
+      "loss": 0.7714,
+      "step": 225
+    },
+    {
+      "epoch": 0.9218436873747495,
+      "grad_norm": 1.034933090209961,
+      "learning_rate": 2.8575446864859115e-05,
+      "loss": 0.8256,
+      "step": 230
+    },
+    {
+      "epoch": 0.9418837675350702,
+      "grad_norm": 1.0317702293395996,
+      "learning_rate": 2.8489829077114644e-05,
+      "loss": 0.7881,
+      "step": 235
+    },
+    {
+      "epoch": 0.9619238476953907,
+      "grad_norm": 1.095956802368164,
+      "learning_rate": 2.8401848971010376e-05,
+      "loss": 0.7371,
+      "step": 240
+    },
+    {
+      "epoch": 0.9819639278557114,
+      "grad_norm": 1.2121639251708984,
+      "learning_rate": 2.8311521953488738e-05,
+      "loss": 0.7487,
+      "step": 245
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.322547197341919,
+      "learning_rate": 2.8218863842479712e-05,
+      "loss": 0.7738,
+      "step": 250
+    },
+    {
+      "epoch": 1.0200400801603207,
+      "grad_norm": 1.5393916368484497,
+      "learning_rate": 2.812389086413084e-05,
+      "loss": 0.6675,
+      "step": 255
+    },
+    {
+      "epoch": 1.0400801603206413,
+      "grad_norm": 1.0528851747512817,
+      "learning_rate": 2.8026619649965703e-05,
+      "loss": 0.6557,
+      "step": 260
+    },
+    {
+      "epoch": 1.060120240480962,
+      "grad_norm": 1.1835638284683228,
+      "learning_rate": 2.7927067233971448e-05,
+      "loss": 0.6464,
+      "step": 265
+    },
+    {
+      "epoch": 1.0801603206412826,
+      "grad_norm": 1.106939435005188,
+      "learning_rate": 2.7825251049615814e-05,
+      "loss": 0.6473,
+      "step": 270
+    },
+    {
+      "epoch": 1.1002004008016033,
+      "grad_norm": 1.1631025075912476,
+      "learning_rate": 2.772118892679422e-05,
+      "loss": 0.6292,
+      "step": 275
+    },
+    {
+      "epoch": 1.1202404809619237,
+      "grad_norm": 1.2355154752731323,
+      "learning_rate": 2.761489908870741e-05,
+      "loss": 0.6232,
+      "step": 280
+    },
+    {
+      "epoch": 1.1402805611222444,
+      "grad_norm": 1.293797492980957,
+      "learning_rate": 2.7506400148670242e-05,
+      "loss": 0.6094,
+      "step": 285
+    },
+    {
+      "epoch": 1.160320641282565,
+      "grad_norm": 1.3766396045684814,
+      "learning_rate": 2.7395711106852152e-05,
+      "loss": 0.5938,
+      "step": 290
+    },
+    {
+      "epoch": 1.1803607214428857,
+      "grad_norm": 1.0483698844909668,
+      "learning_rate": 2.7282851346949854e-05,
+      "loss": 0.6614,
+      "step": 295
+    },
+    {
+      "epoch": 1.2004008016032064,
+      "grad_norm": 1.236453890800476,
+      "learning_rate": 2.7167840632792935e-05,
+      "loss": 0.6253,
+      "step": 300
+    },
+    {
+      "epoch": 1.220440881763527,
+      "grad_norm": 1.3493587970733643,
+      "learning_rate": 2.705069910488281e-05,
+      "loss": 0.605,
+      "step": 305
+    },
+    {
+      "epoch": 1.2404809619238477,
+      "grad_norm": 1.5181005001068115,
+      "learning_rate": 2.6931447276865756e-05,
+      "loss": 0.5845,
+      "step": 310
+    },
+    {
+      "epoch": 1.2605210420841684,
+      "grad_norm": 1.3039430379867554,
+      "learning_rate": 2.6810106031940605e-05,
+      "loss": 0.5867,
+      "step": 315
+    },
+    {
+      "epoch": 1.280561122244489,
+      "grad_norm": 1.1483924388885498,
+      "learning_rate": 2.668669661920169e-05,
+      "loss": 0.6002,
+      "step": 320
+    },
+    {
+      "epoch": 1.3006012024048097,
+      "grad_norm": 1.1646347045898438,
+      "learning_rate": 2.656124064991776e-05,
+      "loss": 0.5763,
+      "step": 325
+    },
+    {
+      "epoch": 1.3206412825651301,
+      "grad_norm": 1.258243203163147,
+      "learning_rate": 2.6433760093747423e-05,
+      "loss": 0.5848,
+      "step": 330
+    },
+    {
+      "epoch": 1.340681362725451,
+      "grad_norm": 1.2748278379440308,
+      "learning_rate": 2.630427727489188e-05,
+      "loss": 0.5547,
+      "step": 335
+    },
+    {
+      "epoch": 1.3607214428857715,
+      "grad_norm": 1.1411411762237549,
+      "learning_rate": 2.6172814868185514e-05,
+      "loss": 0.582,
+      "step": 340
+    },
+    {
+      "epoch": 1.3807615230460921,
+      "grad_norm": 1.139147400856018,
+      "learning_rate": 2.6039395895125126e-05,
+      "loss": 0.5762,
+      "step": 345
+    },
+    {
+      "epoch": 1.4008016032064128,
+      "grad_norm": 1.3767995834350586,
+      "learning_rate": 2.5904043719838454e-05,
+      "loss": 0.5527,
+      "step": 350
+    },
+    {
+      "epoch": 1.4208416833667334,
+      "grad_norm": 1.2413747310638428,
+      "learning_rate": 2.5766782044992658e-05,
+      "loss": 0.5476,
+      "step": 355
+    },
+    {
+      "epoch": 1.440881763527054,
+      "grad_norm": 1.324476957321167,
+      "learning_rate": 2.5627634907643557e-05,
+      "loss": 0.5555,
+      "step": 360
+    },
+    {
+      "epoch": 1.4609218436873748,
+      "grad_norm": 1.1200647354125977,
+      "learning_rate": 2.548662667502631e-05,
+      "loss": 0.5167,
+      "step": 365
+    },
+    {
+      "epoch": 1.4809619238476954,
+      "grad_norm": 1.2644788026809692,
+      "learning_rate": 2.534378204028824e-05,
+      "loss": 0.5192,
+      "step": 370
+    },
+    {
+      "epoch": 1.5010020040080159,
+      "grad_norm": 1.2037605047225952,
+      "learning_rate": 2.5199126018164617e-05,
+      "loss": 0.4919,
+      "step": 375
+    },
+    {
+      "epoch": 1.5210420841683367,
+      "grad_norm": 1.1925784349441528,
+      "learning_rate": 2.5052683940598123e-05,
+      "loss": 0.4958,
+      "step": 380
+    },
+    {
+      "epoch": 1.5410821643286572,
+      "grad_norm": 1.315958857536316,
+      "learning_rate": 2.4904481452302746e-05,
+      "loss": 0.5054,
+      "step": 385
+    },
+    {
+      "epoch": 1.561122244488978,
+      "grad_norm": 1.2814106941223145,
+      "learning_rate": 2.4754544506272933e-05,
+      "loss": 0.5003,
+      "step": 390
+    },
+    {
+      "epoch": 1.5811623246492985,
+      "grad_norm": 1.369860053062439,
+      "learning_rate": 2.4602899359238723e-05,
+      "loss": 0.5134,
+      "step": 395
+    },
+    {
+      "epoch": 1.6012024048096194,
+      "grad_norm": 1.2755391597747803,
+      "learning_rate": 2.4449572567067732e-05,
+      "loss": 0.4822,
+      "step": 400
+    },
+    {
+      "epoch": 1.6212424849699398,
+      "grad_norm": 1.3505276441574097,
+      "learning_rate": 2.4294590980114694e-05,
+      "loss": 0.4356,
+      "step": 405
+    },
+    {
+      "epoch": 1.6412825651302605,
+      "grad_norm": 1.3664660453796387,
+      "learning_rate": 2.413798173851951e-05,
+      "loss": 0.5066,
+      "step": 410
+    },
+    {
+      "epoch": 1.6613226452905812,
+      "grad_norm": 1.2794901132583618,
+      "learning_rate": 2.397977226745448e-05,
+      "loss": 0.4646,
+      "step": 415
+    },
+    {
+      "epoch": 1.6813627254509018,
+      "grad_norm": 1.4925967454910278,
+      "learning_rate": 2.3819990272321653e-05,
+      "loss": 0.453,
+      "step": 420
+    },
+    {
+      "epoch": 1.7014028056112225,
+      "grad_norm": 1.2440918684005737,
+      "learning_rate": 2.3658663733901106e-05,
+      "loss": 0.4694,
+      "step": 425
+    },
+    {
+      "epoch": 1.7214428857715431,
+      "grad_norm": 1.319008231163025,
+      "learning_rate": 2.3495820903450987e-05,
+      "loss": 0.4396,
+      "step": 430
+    },
+    {
+      "epoch": 1.7414829659318638,
+      "grad_norm": 1.2664976119995117,
+      "learning_rate": 2.33314902977602e-05,
+      "loss": 0.4475,
+      "step": 435
+    },
+    {
+      "epoch": 1.7615230460921842,
+      "grad_norm": 1.187595009803772,
+      "learning_rate": 2.3165700694154594e-05,
+      "loss": 0.4437,
+      "step": 440
+    },
+    {
+      "epoch": 1.7815631262525051,
+      "grad_norm": 1.291571021080017,
+      "learning_rate": 2.2998481125457487e-05,
+      "loss": 0.4929,
+      "step": 445
+    },
+    {
+      "epoch": 1.8016032064128256,
+      "grad_norm": 1.313747763633728,
+      "learning_rate": 2.2829860874905553e-05,
+      "loss": 0.4326,
+      "step": 450
+    },
+    {
+      "epoch": 1.8216432865731464,
+      "grad_norm": 1.4631640911102295,
+      "learning_rate": 2.2659869471020705e-05,
+      "loss": 0.457,
+      "step": 455
+    },
+    {
+      "epoch": 1.8416833667334669,
+      "grad_norm": 1.1607767343521118,
+      "learning_rate": 2.2488536682439163e-05,
+      "loss": 0.4325,
+      "step": 460
+    },
+    {
+      "epoch": 1.8617234468937875,
+      "grad_norm": 1.2916303873062134,
+      "learning_rate": 2.2315892512698396e-05,
+      "loss": 0.4441,
+      "step": 465
+    },
+    {
+      "epoch": 1.8817635270541082,
+      "grad_norm": 1.101718544960022,
+      "learning_rate": 2.2141967194982938e-05,
+      "loss": 0.4034,
+      "step": 470
+    },
+    {
+      "epoch": 1.9018036072144289,
+      "grad_norm": 1.2354484796524048,
+      "learning_rate": 2.1966791186830016e-05,
+      "loss": 0.4261,
+      "step": 475
+    },
+    {
+      "epoch": 1.9218436873747495,
+      "grad_norm": 1.177369475364685,
+      "learning_rate": 2.1790395164795857e-05,
+      "loss": 0.4472,
+      "step": 480
+    },
+    {
+      "epoch": 1.9418837675350702,
+      "grad_norm": 1.5405458211898804,
+      "learning_rate": 2.1612810019083684e-05,
+      "loss": 0.357,
+      "step": 485
+    },
+    {
+      "epoch": 1.9619238476953909,
+      "grad_norm": 1.3290022611618042,
+      "learning_rate": 2.143406684813422e-05,
+      "loss": 0.391,
+      "step": 490
+    },
+    {
+      "epoch": 1.9819639278557113,
+      "grad_norm": 1.2326915264129639,
+      "learning_rate": 2.1254196953179833e-05,
+      "loss": 0.382,
+      "step": 495
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.7619742155075073,
+      "learning_rate": 2.107323183276308e-05,
+      "loss": 0.3988,
+      "step": 500
+    },
+    {
+      "epoch": 2.0200400801603204,
+      "grad_norm": 1.3173561096191406,
+      "learning_rate": 2.089120317722072e-05,
+      "loss": 0.3158,
+      "step": 505
+    },
+    {
+      "epoch": 2.0400801603206413,
+      "grad_norm": 1.3548692464828491,
+      "learning_rate": 2.070814286313419e-05,
+      "loss": 0.3485,
+      "step": 510
+    },
+    {
+      "epoch": 2.0601202404809618,
+      "grad_norm": 1.282515287399292,
+      "learning_rate": 2.052408294774739e-05,
+      "loss": 0.3211,
+      "step": 515
+    },
+    {
+      "epoch": 2.0801603206412826,
+      "grad_norm": 1.3488065004348755,
+      "learning_rate": 2.0339055663352874e-05,
+      "loss": 0.2973,
+      "step": 520
+    },
+    {
+      "epoch": 2.100200400801603,
+      "grad_norm": 1.3998934030532837,
+      "learning_rate": 2.0153093411647404e-05,
+      "loss": 0.2996,
+      "step": 525
+    },
+    {
+      "epoch": 2.120240480961924,
+      "grad_norm": 1.321966290473938,
+      "learning_rate": 1.996622875805778e-05,
+      "loss": 0.2758,
+      "step": 530
+    },
+    {
+      "epoch": 2.1402805611222444,
+      "grad_norm": 1.3257852792739868,
+      "learning_rate": 1.977849442603805e-05,
+      "loss": 0.2754,
+      "step": 535
+    },
+    {
+      "epoch": 2.1603206412825653,
+      "grad_norm": 1.4227349758148193,
+      "learning_rate": 1.9589923291339045e-05,
+      "loss": 0.3134,
+      "step": 540
+    },
+    {
+      "epoch": 2.1803607214428857,
+      "grad_norm": 1.3071156740188599,
+      "learning_rate": 1.940054837625118e-05,
+      "loss": 0.3571,
+      "step": 545
+    },
+    {
+      "epoch": 2.2004008016032066,
+      "grad_norm": 1.237615942955017,
+      "learning_rate": 1.9210402843821683e-05,
+      "loss": 0.3019,
+      "step": 550
+    },
+    {
+      "epoch": 2.220440881763527,
+      "grad_norm": 1.3811471462249756,
+      "learning_rate": 1.9019519992047134e-05,
+      "loss": 0.2971,
+      "step": 555
+    },
+    {
+      "epoch": 2.2404809619238475,
+      "grad_norm": 1.5507045984268188,
+      "learning_rate": 1.8827933248042336e-05,
+      "loss": 0.2513,
+      "step": 560
+    },
+    {
+      "epoch": 2.2605210420841684,
+      "grad_norm": 1.3488818407058716,
+      "learning_rate": 1.8635676162186633e-05,
+      "loss": 0.2809,
+      "step": 565
+    },
+    {
+      "epoch": 2.280561122244489,
+      "grad_norm": 1.510488748550415,
+      "learning_rate": 1.8442782402248615e-05,
+      "loss": 0.2788,
+      "step": 570
+    },
+    {
+      "epoch": 2.3006012024048097,
+      "grad_norm": 1.286403775215149,
+      "learning_rate": 1.8249285747490273e-05,
+      "loss": 0.2931,
+      "step": 575
+    },
+    {
+      "epoch": 2.32064128256513,
+      "grad_norm": 1.2511402368545532,
+      "learning_rate": 1.8055220082751627e-05,
+      "loss": 0.3071,
+      "step": 580
+    },
+    {
+      "epoch": 2.340681362725451,
+      "grad_norm": 1.2885857820510864,
+      "learning_rate": 1.7860619392516845e-05,
+      "loss": 0.2775,
+      "step": 585
+    },
+    {
+      "epoch": 2.3607214428857715,
+      "grad_norm": 1.4099708795547485,
+      "learning_rate": 1.7665517754962965e-05,
+      "loss": 0.28,
+      "step": 590
+    },
+    {
+      "epoch": 2.3807615230460923,
+      "grad_norm": 1.2045342922210693,
+      "learning_rate": 1.7469949335992147e-05,
+      "loss": 0.315,
+      "step": 595
+    },
+    {
+      "epoch": 2.400801603206413,
+      "grad_norm": 1.3286011219024658,
+      "learning_rate": 1.727394838324859e-05,
+      "loss": 0.2738,
+      "step": 600
+    },
+    {
+      "epoch": 2.420841683366733,
+      "grad_norm": 1.28937566280365,
+      "learning_rate": 1.707754922012118e-05,
+      "loss": 0.3019,
+      "step": 605
+    },
+    {
+      "epoch": 2.440881763527054,
+      "grad_norm": 1.2608283758163452,
+      "learning_rate": 1.6880786239732782e-05,
+      "loss": 0.2668,
+      "step": 610
+    },
+    {
+      "epoch": 2.460921843687375,
+      "grad_norm": 1.3089014291763306,
+      "learning_rate": 1.668369389891738e-05,
+      "loss": 0.2528,
+      "step": 615
+    },
+    {
+      "epoch": 2.4809619238476954,
+      "grad_norm": 1.657732367515564,
+      "learning_rate": 1.6486306712186073e-05,
+      "loss": 0.2815,
+      "step": 620
+    },
+    {
+      "epoch": 2.501002004008016,
+      "grad_norm": 1.4100173711776733,
+      "learning_rate": 1.6288659245682913e-05,
+      "loss": 0.2759,
+      "step": 625
+    },
+    {
+      "epoch": 2.5210420841683367,
+      "grad_norm": 1.3417761325836182,
+      "learning_rate": 1.6090786111131763e-05,
+      "loss": 0.2785,
+      "step": 630
+    },
+    {
+      "epoch": 2.541082164328657,
+      "grad_norm": 1.2850613594055176,
+      "learning_rate": 1.589272195977512e-05,
+      "loss": 0.2453,
+      "step": 635
+    },
+    {
+      "epoch": 2.561122244488978,
+      "grad_norm": 1.3684494495391846,
+      "learning_rate": 1.5694501476306057e-05,
+      "loss": 0.2607,
+      "step": 640
+    },
+    {
+      "epoch": 2.5811623246492985,
+      "grad_norm": 1.1747251749038696,
+      "learning_rate": 1.5496159372794298e-05,
+      "loss": 0.2634,
+      "step": 645
+    },
+    {
+      "epoch": 2.6012024048096194,
+      "grad_norm": 1.3761392831802368,
+      "learning_rate": 1.529773038260747e-05,
+      "loss": 0.2605,
+      "step": 650
+    },
+    {
+      "epoch": 2.62124248496994,
+      "grad_norm": 1.1896741390228271,
+      "learning_rate": 1.5099249254328666e-05,
+      "loss": 0.2611,
+      "step": 655
+    },
+    {
+      "epoch": 2.6412825651302603,
+      "grad_norm": 1.225896954536438,
+      "learning_rate": 1.490075074567134e-05,
+      "loss": 0.2228,
+      "step": 660
+    },
+    {
+      "epoch": 2.661322645290581,
+      "grad_norm": 1.2646639347076416,
+      "learning_rate": 1.4702269617392537e-05,
+      "loss": 0.2309,
+      "step": 665
+    },
+    {
+      "epoch": 2.681362725450902,
+      "grad_norm": 1.2962318658828735,
+      "learning_rate": 1.4503840627205708e-05,
+      "loss": 0.2572,
+      "step": 670
+    },
+    {
+      "epoch": 2.7014028056112225,
+      "grad_norm": 1.3285491466522217,
+      "learning_rate": 1.4305498523693944e-05,
+      "loss": 0.2195,
+      "step": 675
+    },
+    {
+      "epoch": 2.721442885771543,
+      "grad_norm": 1.3507797718048096,
+      "learning_rate": 1.4107278040224886e-05,
+      "loss": 0.2297,
+      "step": 680
+    },
+    {
+      "epoch": 2.741482965931864,
+      "grad_norm": 1.330955147743225,
+      "learning_rate": 1.3909213888868244e-05,
+      "loss": 0.2366,
+      "step": 685
+    },
+    {
+      "epoch": 2.7615230460921842,
+      "grad_norm": 1.2785532474517822,
+      "learning_rate": 1.3711340754317088e-05,
+      "loss": 0.2169,
+      "step": 690
+    },
+    {
+      "epoch": 2.781563126252505,
+      "grad_norm": 1.2851275205612183,
+      "learning_rate": 1.3513693287813932e-05,
+      "loss": 0.2343,
+      "step": 695
+    },
+    {
+      "epoch": 2.8016032064128256,
+      "grad_norm": 1.5116199254989624,
+      "learning_rate": 1.3316306101082625e-05,
+      "loss": 0.2222,
+      "step": 700
+    },
+    {
+      "epoch": 2.8216432865731464,
+      "grad_norm": 1.1207387447357178,
+      "learning_rate": 1.3119213760267222e-05,
+      "loss": 0.2244,
+      "step": 705
+    },
+    {
+      "epoch": 2.841683366733467,
+      "grad_norm": 1.202231764793396,
+      "learning_rate": 1.2922450779878824e-05,
+      "loss": 0.238,
+      "step": 710
+    },
+    {
+      "epoch": 2.8617234468937873,
+      "grad_norm": 1.4184461832046509,
+      "learning_rate": 1.2726051616751415e-05,
+      "loss": 0.2129,
+      "step": 715
+    },
+    {
+      "epoch": 2.881763527054108,
+      "grad_norm": 1.2793796062469482,
+      "learning_rate": 1.253005066400786e-05,
+      "loss": 0.2107,
+      "step": 720
+    },
+    {
+      "epoch": 2.901803607214429,
+      "grad_norm": 1.4120863676071167,
+      "learning_rate": 1.2334482245037041e-05,
+      "loss": 0.2185,
+      "step": 725
+    },
+    {
+      "epoch": 2.9218436873747495,
+      "grad_norm": 1.1317527294158936,
+      "learning_rate": 1.2139380607483156e-05,
+      "loss": 0.2274,
+      "step": 730
+    },
+    {
+      "epoch": 2.94188376753507,
+      "grad_norm": 1.3882746696472168,
+      "learning_rate": 1.1944779917248379e-05,
+      "loss": 0.2226,
+      "step": 735
+    },
+    {
+      "epoch": 2.961923847695391,
+      "grad_norm": 1.3538494110107422,
+      "learning_rate": 1.1750714252509733e-05,
+      "loss": 0.2144,
+      "step": 740
+    },
+    {
+      "epoch": 2.9819639278557113,
+      "grad_norm": 1.2715141773223877,
+      "learning_rate": 1.1557217597751391e-05,
+      "loss": 0.1835,
+      "step": 745
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.9165490865707397,
+      "learning_rate": 1.1364323837813377e-05,
+      "loss": 0.2478,
+      "step": 750
+    },
+    {
+      "epoch": 3.0200400801603204,
+      "grad_norm": 1.1404417753219604,
+      "learning_rate": 1.1172066751957671e-05,
+      "loss": 0.1838,
+      "step": 755
+    },
+    {
+      "epoch": 3.0400801603206413,
+      "grad_norm": 1.239406943321228,
+      "learning_rate": 1.0980480007952866e-05,
+      "loss": 0.1478,
+      "step": 760
+    },
+    {
+      "epoch": 3.0601202404809618,
+      "grad_norm": 1.1695140600204468,
+      "learning_rate": 1.0789597156178318e-05,
+      "loss": 0.1623,
+      "step": 765
+    },
+    {
+      "epoch": 3.0801603206412826,
+      "grad_norm": 1.2266541719436646,
+      "learning_rate": 1.0599451623748823e-05,
+      "loss": 0.1702,
+      "step": 770
+    },
+    {
+      "epoch": 3.100200400801603,
+      "grad_norm": 1.154406189918518,
+      "learning_rate": 1.0410076708660959e-05,
+      "loss": 0.1585,
+      "step": 775
+    },
+    {
+      "epoch": 3.120240480961924,
+      "grad_norm": 1.3153400421142578,
+      "learning_rate": 1.0221505573961948e-05,
+      "loss": 0.146,
+      "step": 780
+    },
+    {
+      "epoch": 3.1402805611222444,
+      "grad_norm": 1.1898683309555054,
+      "learning_rate": 1.0033771241942221e-05,
+      "loss": 0.1324,
+      "step": 785
+    },
+    {
+      "epoch": 3.1603206412825653,
+      "grad_norm": 1.3161805868148804,
+      "learning_rate": 9.846906588352599e-06,
+      "loss": 0.1508,
+      "step": 790
+    },
+    {
+      "epoch": 3.1803607214428857,
+      "grad_norm": 1.435442328453064,
+      "learning_rate": 9.660944336647129e-06,
+      "loss": 0.1591,
+      "step": 795
+    },
+    {
+      "epoch": 3.2004008016032066,
+      "grad_norm": 1.415054202079773,
+      "learning_rate": 9.475917052252616e-06,
+      "loss": 0.158,
+      "step": 800
+    },
+    {
+      "epoch": 3.220440881763527,
+      "grad_norm": 1.1507920026779175,
+      "learning_rate": 9.291857136865814e-06,
+      "loss": 0.1514,
+      "step": 805
+    },
+    {
+      "epoch": 3.2404809619238475,
+      "grad_norm": 1.2796574831008911,
+      "learning_rate": 9.10879682277928e-06,
+      "loss": 0.1529,
+      "step": 810
+    },
+    {
+      "epoch": 3.2605210420841684,
+      "grad_norm": 1.2656141519546509,
+      "learning_rate": 8.926768167236924e-06,
+      "loss": 0.1402,
+      "step": 815
+    },
+    {
+      "epoch": 3.280561122244489,
+      "grad_norm": 1.0473889112472534,
+      "learning_rate": 8.745803046820168e-06,
+      "loss": 0.1807,
+      "step": 820
+    },
+    {
+      "epoch": 3.3006012024048097,
+      "grad_norm": 0.9760593771934509,
+      "learning_rate": 8.565933151865785e-06,
+      "loss": 0.1514,
+      "step": 825
+    },
+    {
+      "epoch": 3.32064128256513,
+      "grad_norm": 1.1415501832962036,
+      "learning_rate": 8.387189980916318e-06,
+      "loss": 0.1321,
+      "step": 830
+    },
+    {
+      "epoch": 3.340681362725451,
+      "grad_norm": 1.1905689239501953,
+      "learning_rate": 8.20960483520414e-06,
+      "loss": 0.1422,
+      "step": 835
+    },
+    {
+      "epoch": 3.3607214428857715,
+      "grad_norm": 1.0787945985794067,
+      "learning_rate": 8.033208813169986e-06,
+      "loss": 0.1378,
+      "step": 840
+    },
+    {
+      "epoch": 3.3807615230460923,
+      "grad_norm": 1.1386911869049072,
+      "learning_rate": 7.858032805017063e-06,
+      "loss": 0.1244,
+      "step": 845
+    },
+    {
+      "epoch": 3.400801603206413,
+      "grad_norm": 1.272212028503418,
+      "learning_rate": 7.684107487301606e-06,
+      "loss": 0.1422,
+      "step": 850
+    },
+    {
+      "epoch": 3.420841683366733,
+      "grad_norm": 1.29020094871521,
+      "learning_rate": 7.511463317560839e-06,
+      "loss": 0.1379,
+      "step": 855
+    },
+    {
+      "epoch": 3.440881763527054,
+      "grad_norm": 1.1315019130706787,
+      "learning_rate": 7.340130528979295e-06,
+      "loss": 0.1537,
+      "step": 860
+    },
+    {
+      "epoch": 3.460921843687375,
+      "grad_norm": 1.2238397598266602,
+      "learning_rate": 7.17013912509445e-06,
+      "loss": 0.1315,
+      "step": 865
+    },
+    {
+      "epoch": 3.4809619238476954,
+      "grad_norm": 1.290061593055725,
+      "learning_rate": 7.001518874542514e-06,
+      "loss": 0.162,
+      "step": 870
+    },
+    {
+      "epoch": 3.501002004008016,
+      "grad_norm": 1.1427054405212402,
+      "learning_rate": 6.834299305845414e-06,
+      "loss": 0.1381,
+      "step": 875
+    },
+    {
+      "epoch": 3.5210420841683367,
+      "grad_norm": 1.1416689157485962,
+      "learning_rate": 6.668509702239801e-06,
+      "loss": 0.1559,
+      "step": 880
+    },
+    {
+      "epoch": 3.541082164328657,
+      "grad_norm": 1.6092044115066528,
+      "learning_rate": 6.504179096549011e-06,
+      "loss": 0.1455,
+      "step": 885
+    },
+    {
+      "epoch": 3.561122244488978,
+      "grad_norm": 1.2860769033432007,
+      "learning_rate": 6.3413362660988934e-06,
+      "loss": 0.1379,
+      "step": 890
+    },
+    {
+      "epoch": 3.5811623246492985,
+      "grad_norm": 1.0260988473892212,
+      "learning_rate": 6.18000972767835e-06,
+      "loss": 0.1305,
+      "step": 895
+    },
+    {
+      "epoch": 3.6012024048096194,
+      "grad_norm": 1.1337883472442627,
+      "learning_rate": 6.0202277325455246e-06,
+      "loss": 0.1439,
+      "step": 900
+    },
+    {
+      "epoch": 3.62124248496994,
+      "grad_norm": 1.0394679307937622,
+      "learning_rate": 5.862018261480494e-06,
+      "loss": 0.1321,
+      "step": 905
+    },
+    {
+      "epoch": 3.6412825651302603,
+      "grad_norm": 1.0980677604675293,
+      "learning_rate": 5.7054090198853055e-06,
+      "loss": 0.1371,
+      "step": 910
+    },
+    {
+      "epoch": 3.661322645290581,
+      "grad_norm": 0.9747451543807983,
+      "learning_rate": 5.550427432932272e-06,
+      "loss": 0.1381,
+      "step": 915
+    },
+    {
+      "epoch": 3.681362725450902,
+      "grad_norm": 0.9877727627754211,
+      "learning_rate": 5.397100640761277e-06,
+      "loss": 0.1188,
+      "step": 920
+    },
+    {
+      "epoch": 3.7014028056112225,
+      "grad_norm": 1.2609772682189941,
+      "learning_rate": 5.24545549372707e-06,
+      "loss": 0.1424,
+      "step": 925
+    },
+    {
+      "epoch": 3.721442885771543,
+      "grad_norm": 1.0329722166061401,
+      "learning_rate": 5.0955185476972566e-06,
+      "loss": 0.124,
+      "step": 930
+    },
+    {
+      "epoch": 3.741482965931864,
+      "grad_norm": 1.0099831819534302,
+      "learning_rate": 4.9473160594018815e-06,
+      "loss": 0.1221,
+      "step": 935
+    },
+    {
+      "epoch": 3.7615230460921842,
+      "grad_norm": 1.1363469362258911,
+      "learning_rate": 4.8008739818353824e-06,
+      "loss": 0.1353,
+      "step": 940
+    },
+    {
+      "epoch": 3.781563126252505,
+      "grad_norm": 1.0320320129394531,
+      "learning_rate": 4.656217959711761e-06,
+      "loss": 0.1397,
+      "step": 945
+    },
+    {
+      "epoch": 3.8016032064128256,
+      "grad_norm": 1.3003877401351929,
+      "learning_rate": 4.513373324973691e-06,
+      "loss": 0.1284,
+      "step": 950
+    },
+    {
+      "epoch": 3.8216432865731464,
+      "grad_norm": 1.0216275453567505,
+      "learning_rate": 4.372365092356445e-06,
+      "loss": 0.1049,
+      "step": 955
+    },
+    {
+      "epoch": 3.841683366733467,
+      "grad_norm": 1.0927823781967163,
+      "learning_rate": 4.233217955007346e-06,
+      "loss": 0.1245,
+      "step": 960
+    },
+    {
+      "epoch": 3.8617234468937873,
+      "grad_norm": 1.0101195573806763,
+      "learning_rate": 4.0959562801615445e-06,
+      "loss": 0.1274,
+      "step": 965
+    },
+    {
+      "epoch": 3.881763527054108,
+      "grad_norm": 1.254036784172058,
+      "learning_rate": 3.960604104874871e-06,
+      "loss": 0.1392,
+      "step": 970
+    },
+    {
+      "epoch": 3.901803607214429,
+      "grad_norm": 1.62558114528656,
+      "learning_rate": 3.8271851318144865e-06,
+      "loss": 0.1318,
+      "step": 975
+    },
+    {
+      "epoch": 3.9218436873747495,
+      "grad_norm": 1.004096269607544,
+      "learning_rate": 3.695722725108121e-06,
+      "loss": 0.1221,
+      "step": 980
+    },
+    {
+      "epoch": 3.94188376753507,
+      "grad_norm": 1.1071776151657104,
+      "learning_rate": 3.566239906252576e-06,
+      "loss": 0.134,
+      "step": 985
+    },
+    {
+      "epoch": 3.961923847695391,
+      "grad_norm": 1.0152360200881958,
+      "learning_rate": 3.438759350082243e-06,
+      "loss": 0.1219,
+      "step": 990
+    },
+    {
+      "epoch": 3.9819639278557113,
+      "grad_norm": 1.0832067728042603,
+      "learning_rate": 3.31330338079831e-06,
+      "loss": 0.1198,
+      "step": 995
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.064515471458435,
+      "learning_rate": 3.1898939680593976e-06,
+      "loss": 0.1227,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1250,
+  "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": 1.337039639046783e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

50_128_e5_3e-5/checkpoint-1000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61c16f5cfa671126a0a7b100dbd969a49a07ef964c5bc306796be064502b8d67
+size 7736

50_128_e5_3e-5/checkpoint-1000/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)

50_128_e5_3e-5/checkpoint-1250/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

50_128_e5_3e-5/checkpoint-1250/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": [
+    "o_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

50_128_e5_3e-5/checkpoint-1250/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d412ceaa7b46c812d083a9efde6c92e5e68e44b5c8dc5806a2f07b033da474a
+size 791751704

50_128_e5_3e-5/checkpoint-1250/latest

@@ -0,0 +1 @@
+global_step1250

50_128_e5_3e-5/checkpoint-1250/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed17ffca447fb5bb6b9535cf97d6119eafa3a0b7b4718a47e49fe83827e7bd79
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e062500084b2558dbaeffdf82daa24b8950faaedc0e768178415c6c0aeececa
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cef641404c09da6ebbe01c1ae392d4121dc4560a9627ef05a719dbd8c3261000
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb58e5529a8bd4248ebe69c954050fec9f9e8aace7f8f865c2ccacd0f996acaf
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1b2d6d463f047d6c272afd5d2cb856205f8ebe89f0955a5dfadbfe93ef06414
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1739983338b5b0efb72ead4f64d3d4201d185aeae8578aa2ada176ca0afa6a0c
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b656757d58af8a0adcea26edcdeda2ca265638fb1d3627c370236d01d6f5ee98
+size 15920

50_128_e5_3e-5/checkpoint-1250/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c8b8ef5cd1bc96aab3f79c3c6177ff639591d75469a33c0b237e69ad4efb2dc
+size 15920

50_128_e5_3e-5/checkpoint-1250/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc63bac2fa72bb19b007f6650fe43ab7d7b146c5d3f3b18dbb513fa8a5577473
+size 1064

50_128_e5_3e-5/checkpoint-1250/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "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": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

50_128_e5_3e-5/checkpoint-1250/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": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

50_128_e5_3e-5/checkpoint-1250/trainer_state.json

@@ -0,0 +1,1784 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 1250,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02004008016032064,
+      "grad_norm": 0.9890378713607788,
+      "learning_rate": 1.9047619047619047e-06,
+      "loss": 1.4301,
+      "step": 5
+    },
+    {
+      "epoch": 0.04008016032064128,
+      "grad_norm": 0.863508939743042,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 1.4641,
+      "step": 10
+    },
+    {
+      "epoch": 0.06012024048096192,
+      "grad_norm": 0.6787753701210022,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.4197,
+      "step": 15
+    },
+    {
+      "epoch": 0.08016032064128256,
+      "grad_norm": 0.6681222319602966,
+      "learning_rate": 9.047619047619047e-06,
+      "loss": 1.4312,
+      "step": 20
+    },
+    {
+      "epoch": 0.10020040080160321,
+      "grad_norm": 0.5663285255432129,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 1.3627,
+      "step": 25
+    },
+    {
+      "epoch": 0.12024048096192384,
+      "grad_norm": 0.5837746858596802,
+      "learning_rate": 1.380952380952381e-05,
+      "loss": 1.3997,
+      "step": 30
+    },
+    {
+      "epoch": 0.1402805611222445,
+      "grad_norm": 0.5583487153053284,
+      "learning_rate": 1.619047619047619e-05,
+      "loss": 1.3404,
+      "step": 35
+    },
+    {
+      "epoch": 0.16032064128256512,
+      "grad_norm": 0.4913857877254486,
+      "learning_rate": 1.8571428571428572e-05,
+      "loss": 1.3386,
+      "step": 40
+    },
+    {
+      "epoch": 0.18036072144288579,
+      "grad_norm": 0.573939323425293,
+      "learning_rate": 2.095238095238095e-05,
+      "loss": 1.3201,
+      "step": 45
+    },
+    {
+      "epoch": 0.20040080160320642,
+      "grad_norm": 0.508334219455719,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 1.2503,
+      "step": 50
+    },
+    {
+      "epoch": 0.22044088176352705,
+      "grad_norm": 0.5239691734313965,
+      "learning_rate": 2.5714285714285714e-05,
+      "loss": 1.2509,
+      "step": 55
+    },
+    {
+      "epoch": 0.24048096192384769,
+      "grad_norm": 0.7545399069786072,
+      "learning_rate": 2.8095238095238096e-05,
+      "loss": 1.2877,
+      "step": 60
+    },
+    {
+      "epoch": 0.2605210420841683,
+      "grad_norm": 0.4638131260871887,
+      "learning_rate": 2.9999947463720068e-05,
+      "loss": 1.219,
+      "step": 65
+    },
+    {
+      "epoch": 0.280561122244489,
+      "grad_norm": 0.5463809370994568,
+      "learning_rate": 2.9998108732563026e-05,
+      "loss": 1.2131,
+      "step": 70
+    },
+    {
+      "epoch": 0.30060120240480964,
+      "grad_norm": 0.5909023284912109,
+      "learning_rate": 2.9993643555405922e-05,
+      "loss": 1.1617,
+      "step": 75
+    },
+    {
+      "epoch": 0.32064128256513025,
+      "grad_norm": 0.4846280515193939,
+      "learning_rate": 2.9986552714183782e-05,
+      "loss": 1.2422,
+      "step": 80
+    },
+    {
+      "epoch": 0.3406813627254509,
+      "grad_norm": 0.6445205807685852,
+      "learning_rate": 2.9976837450633944e-05,
+      "loss": 1.1748,
+      "step": 85
+    },
+    {
+      "epoch": 0.36072144288577157,
+      "grad_norm": 0.6564697623252869,
+      "learning_rate": 2.996449946607859e-05,
+      "loss": 1.1738,
+      "step": 90
+    },
+    {
+      "epoch": 0.3807615230460922,
+      "grad_norm": 0.7153516411781311,
+      "learning_rate": 2.9949540921126824e-05,
+      "loss": 1.1438,
+      "step": 95
+    },
+    {
+      "epoch": 0.40080160320641284,
+      "grad_norm": 0.6607214212417603,
+      "learning_rate": 2.9931964435296292e-05,
+      "loss": 1.1229,
+      "step": 100
+    },
+    {
+      "epoch": 0.42084168336673344,
+      "grad_norm": 0.6068851351737976,
+      "learning_rate": 2.991177308655447e-05,
+      "loss": 1.1227,
+      "step": 105
+    },
+    {
+      "epoch": 0.4408817635270541,
+      "grad_norm": 0.6417985558509827,
+      "learning_rate": 2.988897041077966e-05,
+      "loss": 1.1068,
+      "step": 110
+    },
+    {
+      "epoch": 0.46092184368737477,
+      "grad_norm": 0.6902046203613281,
+      "learning_rate": 2.9863560401141773e-05,
+      "loss": 1.1214,
+      "step": 115
+    },
+    {
+      "epoch": 0.48096192384769537,
+      "grad_norm": 0.6481944918632507,
+      "learning_rate": 2.9835547507403067e-05,
+      "loss": 1.0765,
+      "step": 120
+    },
+    {
+      "epoch": 0.501002004008016,
+      "grad_norm": 0.7215368747711182,
+      "learning_rate": 2.980493663513891e-05,
+      "loss": 1.0368,
+      "step": 125
+    },
+    {
+      "epoch": 0.5210420841683366,
+      "grad_norm": 0.7066978216171265,
+      "learning_rate": 2.9771733144878706e-05,
+      "loss": 0.9668,
+      "step": 130
+    },
+    {
+      "epoch": 0.5410821643286573,
+      "grad_norm": 0.7489072680473328,
+      "learning_rate": 2.9735942851167202e-05,
+      "loss": 1.0239,
+      "step": 135
+    },
+    {
+      "epoch": 0.561122244488978,
+      "grad_norm": 0.7932925224304199,
+      "learning_rate": 2.9697572021546216e-05,
+      "loss": 0.9795,
+      "step": 140
+    },
+    {
+      "epoch": 0.5811623246492986,
+      "grad_norm": 0.9287444949150085,
+      "learning_rate": 2.9656627375457102e-05,
+      "loss": 0.9915,
+      "step": 145
+    },
+    {
+      "epoch": 0.6012024048096193,
+      "grad_norm": 0.8538782596588135,
+      "learning_rate": 2.961311608306403e-05,
+      "loss": 1.0117,
+      "step": 150
+    },
+    {
+      "epoch": 0.6212424849699398,
+      "grad_norm": 0.773285984992981,
+      "learning_rate": 2.956704576399838e-05,
+      "loss": 0.9697,
+      "step": 155
+    },
+    {
+      "epoch": 0.6412825651302605,
+      "grad_norm": 0.9210479855537415,
+      "learning_rate": 2.9518424486024382e-05,
+      "loss": 0.9589,
+      "step": 160
+    },
+    {
+      "epoch": 0.6613226452905812,
+      "grad_norm": 0.7253521680831909,
+      "learning_rate": 2.9467260763626323e-05,
+      "loss": 0.9559,
+      "step": 165
+    },
+    {
+      "epoch": 0.6813627254509018,
+      "grad_norm": 1.1397440433502197,
+      "learning_rate": 2.9413563556517483e-05,
+      "loss": 0.9394,
+      "step": 170
+    },
+    {
+      "epoch": 0.7014028056112225,
+      "grad_norm": 0.9524412751197815,
+      "learning_rate": 2.935734226807114e-05,
+      "loss": 0.9114,
+      "step": 175
+    },
+    {
+      "epoch": 0.7214428857715431,
+      "grad_norm": 0.8871064186096191,
+      "learning_rate": 2.9298606743673854e-05,
+      "loss": 0.865,
+      "step": 180
+    },
+    {
+      "epoch": 0.7414829659318637,
+      "grad_norm": 0.9255661368370056,
+      "learning_rate": 2.9237367269001362e-05,
+      "loss": 0.9108,
+      "step": 185
+    },
+    {
+      "epoch": 0.7615230460921844,
+      "grad_norm": 0.9735074639320374,
+      "learning_rate": 2.9173634568217366e-05,
+      "loss": 0.8465,
+      "step": 190
+    },
+    {
+      "epoch": 0.781563126252505,
+      "grad_norm": 0.9207878112792969,
+      "learning_rate": 2.9107419802095546e-05,
+      "loss": 0.8566,
+      "step": 195
+    },
+    {
+      "epoch": 0.8016032064128257,
+      "grad_norm": 0.9188950657844543,
+      "learning_rate": 2.9038734566065068e-05,
+      "loss": 0.815,
+      "step": 200
+    },
+    {
+      "epoch": 0.8216432865731463,
+      "grad_norm": 1.1056013107299805,
+      "learning_rate": 2.8967590888180052e-05,
+      "loss": 0.8299,
+      "step": 205
+    },
+    {
+      "epoch": 0.8416833667334669,
+      "grad_norm": 1.0484192371368408,
+      "learning_rate": 2.8894001227013213e-05,
+      "loss": 0.8266,
+      "step": 210
+    },
+    {
+      "epoch": 0.8617234468937875,
+      "grad_norm": 0.9629917144775391,
+      "learning_rate": 2.8817978469474137e-05,
+      "loss": 0.8386,
+      "step": 215
+    },
+    {
+      "epoch": 0.8817635270541082,
+      "grad_norm": 0.9744315147399902,
+      "learning_rate": 2.873953592855255e-05,
+      "loss": 0.7977,
+      "step": 220
+    },
+    {
+      "epoch": 0.9018036072144289,
+      "grad_norm": 1.0901851654052734,
+      "learning_rate": 2.8658687340986962e-05,
+      "loss": 0.7714,
+      "step": 225
+    },
+    {
+      "epoch": 0.9218436873747495,
+      "grad_norm": 1.034933090209961,
+      "learning_rate": 2.8575446864859115e-05,
+      "loss": 0.8256,
+      "step": 230
+    },
+    {
+      "epoch": 0.9418837675350702,
+      "grad_norm": 1.0317702293395996,
+      "learning_rate": 2.8489829077114644e-05,
+      "loss": 0.7881,
+      "step": 235
+    },
+    {
+      "epoch": 0.9619238476953907,
+      "grad_norm": 1.095956802368164,
+      "learning_rate": 2.8401848971010376e-05,
+      "loss": 0.7371,
+      "step": 240
+    },
+    {
+      "epoch": 0.9819639278557114,
+      "grad_norm": 1.2121639251708984,
+      "learning_rate": 2.8311521953488738e-05,
+      "loss": 0.7487,
+      "step": 245
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.322547197341919,
+      "learning_rate": 2.8218863842479712e-05,
+      "loss": 0.7738,
+      "step": 250
+    },
+    {
+      "epoch": 1.0200400801603207,
+      "grad_norm": 1.5393916368484497,
+      "learning_rate": 2.812389086413084e-05,
+      "loss": 0.6675,
+      "step": 255
+    },
+    {
+      "epoch": 1.0400801603206413,
+      "grad_norm": 1.0528851747512817,
+      "learning_rate": 2.8026619649965703e-05,
+      "loss": 0.6557,
+      "step": 260
+    },
+    {
+      "epoch": 1.060120240480962,
+      "grad_norm": 1.1835638284683228,
+      "learning_rate": 2.7927067233971448e-05,
+      "loss": 0.6464,
+      "step": 265
+    },
+    {
+      "epoch": 1.0801603206412826,
+      "grad_norm": 1.106939435005188,
+      "learning_rate": 2.7825251049615814e-05,
+      "loss": 0.6473,
+      "step": 270
+    },
+    {
+      "epoch": 1.1002004008016033,
+      "grad_norm": 1.1631025075912476,
+      "learning_rate": 2.772118892679422e-05,
+      "loss": 0.6292,
+      "step": 275
+    },
+    {
+      "epoch": 1.1202404809619237,
+      "grad_norm": 1.2355154752731323,
+      "learning_rate": 2.761489908870741e-05,
+      "loss": 0.6232,
+      "step": 280
+    },
+    {
+      "epoch": 1.1402805611222444,
+      "grad_norm": 1.293797492980957,
+      "learning_rate": 2.7506400148670242e-05,
+      "loss": 0.6094,
+      "step": 285
+    },
+    {
+      "epoch": 1.160320641282565,
+      "grad_norm": 1.3766396045684814,
+      "learning_rate": 2.7395711106852152e-05,
+      "loss": 0.5938,
+      "step": 290
+    },
+    {
+      "epoch": 1.1803607214428857,
+      "grad_norm": 1.0483698844909668,
+      "learning_rate": 2.7282851346949854e-05,
+      "loss": 0.6614,
+      "step": 295
+    },
+    {
+      "epoch": 1.2004008016032064,
+      "grad_norm": 1.236453890800476,
+      "learning_rate": 2.7167840632792935e-05,
+      "loss": 0.6253,
+      "step": 300
+    },
+    {
+      "epoch": 1.220440881763527,
+      "grad_norm": 1.3493587970733643,
+      "learning_rate": 2.705069910488281e-05,
+      "loss": 0.605,
+      "step": 305
+    },
+    {
+      "epoch": 1.2404809619238477,
+      "grad_norm": 1.5181005001068115,
+      "learning_rate": 2.6931447276865756e-05,
+      "loss": 0.5845,
+      "step": 310
+    },
+    {
+      "epoch": 1.2605210420841684,
+      "grad_norm": 1.3039430379867554,
+      "learning_rate": 2.6810106031940605e-05,
+      "loss": 0.5867,
+      "step": 315
+    },
+    {
+      "epoch": 1.280561122244489,
+      "grad_norm": 1.1483924388885498,
+      "learning_rate": 2.668669661920169e-05,
+      "loss": 0.6002,
+      "step": 320
+    },
+    {
+      "epoch": 1.3006012024048097,
+      "grad_norm": 1.1646347045898438,
+      "learning_rate": 2.656124064991776e-05,
+      "loss": 0.5763,
+      "step": 325
+    },
+    {
+      "epoch": 1.3206412825651301,
+      "grad_norm": 1.258243203163147,
+      "learning_rate": 2.6433760093747423e-05,
+      "loss": 0.5848,
+      "step": 330
+    },
+    {
+      "epoch": 1.340681362725451,
+      "grad_norm": 1.2748278379440308,
+      "learning_rate": 2.630427727489188e-05,
+      "loss": 0.5547,
+      "step": 335
+    },
+    {
+      "epoch": 1.3607214428857715,
+      "grad_norm": 1.1411411762237549,
+      "learning_rate": 2.6172814868185514e-05,
+      "loss": 0.582,
+      "step": 340
+    },
+    {
+      "epoch": 1.3807615230460921,
+      "grad_norm": 1.139147400856018,
+      "learning_rate": 2.6039395895125126e-05,
+      "loss": 0.5762,
+      "step": 345
+    },
+    {
+      "epoch": 1.4008016032064128,
+      "grad_norm": 1.3767995834350586,
+      "learning_rate": 2.5904043719838454e-05,
+      "loss": 0.5527,
+      "step": 350
+    },
+    {
+      "epoch": 1.4208416833667334,
+      "grad_norm": 1.2413747310638428,
+      "learning_rate": 2.5766782044992658e-05,
+      "loss": 0.5476,
+      "step": 355
+    },
+    {
+      "epoch": 1.440881763527054,
+      "grad_norm": 1.324476957321167,
+      "learning_rate": 2.5627634907643557e-05,
+      "loss": 0.5555,
+      "step": 360
+    },
+    {
+      "epoch": 1.4609218436873748,
+      "grad_norm": 1.1200647354125977,
+      "learning_rate": 2.548662667502631e-05,
+      "loss": 0.5167,
+      "step": 365
+    },
+    {
+      "epoch": 1.4809619238476954,
+      "grad_norm": 1.2644788026809692,
+      "learning_rate": 2.534378204028824e-05,
+      "loss": 0.5192,
+      "step": 370
+    },
+    {
+      "epoch": 1.5010020040080159,
+      "grad_norm": 1.2037605047225952,
+      "learning_rate": 2.5199126018164617e-05,
+      "loss": 0.4919,
+      "step": 375
+    },
+    {
+      "epoch": 1.5210420841683367,
+      "grad_norm": 1.1925784349441528,
+      "learning_rate": 2.5052683940598123e-05,
+      "loss": 0.4958,
+      "step": 380
+    },
+    {
+      "epoch": 1.5410821643286572,
+      "grad_norm": 1.315958857536316,
+      "learning_rate": 2.4904481452302746e-05,
+      "loss": 0.5054,
+      "step": 385
+    },
+    {
+      "epoch": 1.561122244488978,
+      "grad_norm": 1.2814106941223145,
+      "learning_rate": 2.4754544506272933e-05,
+      "loss": 0.5003,
+      "step": 390
+    },
+    {
+      "epoch": 1.5811623246492985,
+      "grad_norm": 1.369860053062439,
+      "learning_rate": 2.4602899359238723e-05,
+      "loss": 0.5134,
+      "step": 395
+    },
+    {
+      "epoch": 1.6012024048096194,
+      "grad_norm": 1.2755391597747803,
+      "learning_rate": 2.4449572567067732e-05,
+      "loss": 0.4822,
+      "step": 400
+    },
+    {
+      "epoch": 1.6212424849699398,
+      "grad_norm": 1.3505276441574097,
+      "learning_rate": 2.4294590980114694e-05,
+      "loss": 0.4356,
+      "step": 405
+    },
+    {
+      "epoch": 1.6412825651302605,
+      "grad_norm": 1.3664660453796387,
+      "learning_rate": 2.413798173851951e-05,
+      "loss": 0.5066,
+      "step": 410
+    },
+    {
+      "epoch": 1.6613226452905812,
+      "grad_norm": 1.2794901132583618,
+      "learning_rate": 2.397977226745448e-05,
+      "loss": 0.4646,
+      "step": 415
+    },
+    {
+      "epoch": 1.6813627254509018,
+      "grad_norm": 1.4925967454910278,
+      "learning_rate": 2.3819990272321653e-05,
+      "loss": 0.453,
+      "step": 420
+    },
+    {
+      "epoch": 1.7014028056112225,
+      "grad_norm": 1.2440918684005737,
+      "learning_rate": 2.3658663733901106e-05,
+      "loss": 0.4694,
+      "step": 425
+    },
+    {
+      "epoch": 1.7214428857715431,
+      "grad_norm": 1.319008231163025,
+      "learning_rate": 2.3495820903450987e-05,
+      "loss": 0.4396,
+      "step": 430
+    },
+    {
+      "epoch": 1.7414829659318638,
+      "grad_norm": 1.2664976119995117,
+      "learning_rate": 2.33314902977602e-05,
+      "loss": 0.4475,
+      "step": 435
+    },
+    {
+      "epoch": 1.7615230460921842,
+      "grad_norm": 1.187595009803772,
+      "learning_rate": 2.3165700694154594e-05,
+      "loss": 0.4437,
+      "step": 440
+    },
+    {
+      "epoch": 1.7815631262525051,
+      "grad_norm": 1.291571021080017,
+      "learning_rate": 2.2998481125457487e-05,
+      "loss": 0.4929,
+      "step": 445
+    },
+    {
+      "epoch": 1.8016032064128256,
+      "grad_norm": 1.313747763633728,
+      "learning_rate": 2.2829860874905553e-05,
+      "loss": 0.4326,
+      "step": 450
+    },
+    {
+      "epoch": 1.8216432865731464,
+      "grad_norm": 1.4631640911102295,
+      "learning_rate": 2.2659869471020705e-05,
+      "loss": 0.457,
+      "step": 455
+    },
+    {
+      "epoch": 1.8416833667334669,
+      "grad_norm": 1.1607767343521118,
+      "learning_rate": 2.2488536682439163e-05,
+      "loss": 0.4325,
+      "step": 460
+    },
+    {
+      "epoch": 1.8617234468937875,
+      "grad_norm": 1.2916303873062134,
+      "learning_rate": 2.2315892512698396e-05,
+      "loss": 0.4441,
+      "step": 465
+    },
+    {
+      "epoch": 1.8817635270541082,
+      "grad_norm": 1.101718544960022,
+      "learning_rate": 2.2141967194982938e-05,
+      "loss": 0.4034,
+      "step": 470
+    },
+    {
+      "epoch": 1.9018036072144289,
+      "grad_norm": 1.2354484796524048,
+      "learning_rate": 2.1966791186830016e-05,
+      "loss": 0.4261,
+      "step": 475
+    },
+    {
+      "epoch": 1.9218436873747495,
+      "grad_norm": 1.177369475364685,
+      "learning_rate": 2.1790395164795857e-05,
+      "loss": 0.4472,
+      "step": 480
+    },
+    {
+      "epoch": 1.9418837675350702,
+      "grad_norm": 1.5405458211898804,
+      "learning_rate": 2.1612810019083684e-05,
+      "loss": 0.357,
+      "step": 485
+    },
+    {
+      "epoch": 1.9619238476953909,
+      "grad_norm": 1.3290022611618042,
+      "learning_rate": 2.143406684813422e-05,
+      "loss": 0.391,
+      "step": 490
+    },
+    {
+      "epoch": 1.9819639278557113,
+      "grad_norm": 1.2326915264129639,
+      "learning_rate": 2.1254196953179833e-05,
+      "loss": 0.382,
+      "step": 495
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.7619742155075073,
+      "learning_rate": 2.107323183276308e-05,
+      "loss": 0.3988,
+      "step": 500
+    },
+    {
+      "epoch": 2.0200400801603204,
+      "grad_norm": 1.3173561096191406,
+      "learning_rate": 2.089120317722072e-05,
+      "loss": 0.3158,
+      "step": 505
+    },
+    {
+      "epoch": 2.0400801603206413,
+      "grad_norm": 1.3548692464828491,
+      "learning_rate": 2.070814286313419e-05,
+      "loss": 0.3485,
+      "step": 510
+    },
+    {
+      "epoch": 2.0601202404809618,
+      "grad_norm": 1.282515287399292,
+      "learning_rate": 2.052408294774739e-05,
+      "loss": 0.3211,
+      "step": 515
+    },
+    {
+      "epoch": 2.0801603206412826,
+      "grad_norm": 1.3488065004348755,
+      "learning_rate": 2.0339055663352874e-05,
+      "loss": 0.2973,
+      "step": 520
+    },
+    {
+      "epoch": 2.100200400801603,
+      "grad_norm": 1.3998934030532837,
+      "learning_rate": 2.0153093411647404e-05,
+      "loss": 0.2996,
+      "step": 525
+    },
+    {
+      "epoch": 2.120240480961924,
+      "grad_norm": 1.321966290473938,
+      "learning_rate": 1.996622875805778e-05,
+      "loss": 0.2758,
+      "step": 530
+    },
+    {
+      "epoch": 2.1402805611222444,
+      "grad_norm": 1.3257852792739868,
+      "learning_rate": 1.977849442603805e-05,
+      "loss": 0.2754,
+      "step": 535
+    },
+    {
+      "epoch": 2.1603206412825653,
+      "grad_norm": 1.4227349758148193,
+      "learning_rate": 1.9589923291339045e-05,
+      "loss": 0.3134,
+      "step": 540
+    },
+    {
+      "epoch": 2.1803607214428857,
+      "grad_norm": 1.3071156740188599,
+      "learning_rate": 1.940054837625118e-05,
+      "loss": 0.3571,
+      "step": 545
+    },
+    {
+      "epoch": 2.2004008016032066,
+      "grad_norm": 1.237615942955017,
+      "learning_rate": 1.9210402843821683e-05,
+      "loss": 0.3019,
+      "step": 550
+    },
+    {
+      "epoch": 2.220440881763527,
+      "grad_norm": 1.3811471462249756,
+      "learning_rate": 1.9019519992047134e-05,
+      "loss": 0.2971,
+      "step": 555
+    },
+    {
+      "epoch": 2.2404809619238475,
+      "grad_norm": 1.5507045984268188,
+      "learning_rate": 1.8827933248042336e-05,
+      "loss": 0.2513,
+      "step": 560
+    },
+    {
+      "epoch": 2.2605210420841684,
+      "grad_norm": 1.3488818407058716,
+      "learning_rate": 1.8635676162186633e-05,
+      "loss": 0.2809,
+      "step": 565
+    },
+    {
+      "epoch": 2.280561122244489,
+      "grad_norm": 1.510488748550415,
+      "learning_rate": 1.8442782402248615e-05,
+      "loss": 0.2788,
+      "step": 570
+    },
+    {
+      "epoch": 2.3006012024048097,
+      "grad_norm": 1.286403775215149,
+      "learning_rate": 1.8249285747490273e-05,
+      "loss": 0.2931,
+      "step": 575
+    },
+    {
+      "epoch": 2.32064128256513,
+      "grad_norm": 1.2511402368545532,
+      "learning_rate": 1.8055220082751627e-05,
+      "loss": 0.3071,
+      "step": 580
+    },
+    {
+      "epoch": 2.340681362725451,
+      "grad_norm": 1.2885857820510864,
+      "learning_rate": 1.7860619392516845e-05,
+      "loss": 0.2775,
+      "step": 585
+    },
+    {
+      "epoch": 2.3607214428857715,
+      "grad_norm": 1.4099708795547485,
+      "learning_rate": 1.7665517754962965e-05,
+      "loss": 0.28,
+      "step": 590
+    },
+    {
+      "epoch": 2.3807615230460923,
+      "grad_norm": 1.2045342922210693,
+      "learning_rate": 1.7469949335992147e-05,
+      "loss": 0.315,
+      "step": 595
+    },
+    {
+      "epoch": 2.400801603206413,
+      "grad_norm": 1.3286011219024658,
+      "learning_rate": 1.727394838324859e-05,
+      "loss": 0.2738,
+      "step": 600
+    },
+    {
+      "epoch": 2.420841683366733,
+      "grad_norm": 1.28937566280365,
+      "learning_rate": 1.707754922012118e-05,
+      "loss": 0.3019,
+      "step": 605
+    },
+    {
+      "epoch": 2.440881763527054,
+      "grad_norm": 1.2608283758163452,
+      "learning_rate": 1.6880786239732782e-05,
+      "loss": 0.2668,
+      "step": 610
+    },
+    {
+      "epoch": 2.460921843687375,
+      "grad_norm": 1.3089014291763306,
+      "learning_rate": 1.668369389891738e-05,
+      "loss": 0.2528,
+      "step": 615
+    },
+    {
+      "epoch": 2.4809619238476954,
+      "grad_norm": 1.657732367515564,
+      "learning_rate": 1.6486306712186073e-05,
+      "loss": 0.2815,
+      "step": 620
+    },
+    {
+      "epoch": 2.501002004008016,
+      "grad_norm": 1.4100173711776733,
+      "learning_rate": 1.6288659245682913e-05,
+      "loss": 0.2759,
+      "step": 625
+    },
+    {
+      "epoch": 2.5210420841683367,
+      "grad_norm": 1.3417761325836182,
+      "learning_rate": 1.6090786111131763e-05,
+      "loss": 0.2785,
+      "step": 630
+    },
+    {
+      "epoch": 2.541082164328657,
+      "grad_norm": 1.2850613594055176,
+      "learning_rate": 1.589272195977512e-05,
+      "loss": 0.2453,
+      "step": 635
+    },
+    {
+      "epoch": 2.561122244488978,
+      "grad_norm": 1.3684494495391846,
+      "learning_rate": 1.5694501476306057e-05,
+      "loss": 0.2607,
+      "step": 640
+    },
+    {
+      "epoch": 2.5811623246492985,
+      "grad_norm": 1.1747251749038696,
+      "learning_rate": 1.5496159372794298e-05,
+      "loss": 0.2634,
+      "step": 645
+    },
+    {
+      "epoch": 2.6012024048096194,
+      "grad_norm": 1.3761392831802368,
+      "learning_rate": 1.529773038260747e-05,
+      "loss": 0.2605,
+      "step": 650
+    },
+    {
+      "epoch": 2.62124248496994,
+      "grad_norm": 1.1896741390228271,
+      "learning_rate": 1.5099249254328666e-05,
+      "loss": 0.2611,
+      "step": 655
+    },
+    {
+      "epoch": 2.6412825651302603,
+      "grad_norm": 1.225896954536438,
+      "learning_rate": 1.490075074567134e-05,
+      "loss": 0.2228,
+      "step": 660
+    },
+    {
+      "epoch": 2.661322645290581,
+      "grad_norm": 1.2646639347076416,
+      "learning_rate": 1.4702269617392537e-05,
+      "loss": 0.2309,
+      "step": 665
+    },
+    {
+      "epoch": 2.681362725450902,
+      "grad_norm": 1.2962318658828735,
+      "learning_rate": 1.4503840627205708e-05,
+      "loss": 0.2572,
+      "step": 670
+    },
+    {
+      "epoch": 2.7014028056112225,
+      "grad_norm": 1.3285491466522217,
+      "learning_rate": 1.4305498523693944e-05,
+      "loss": 0.2195,
+      "step": 675
+    },
+    {
+      "epoch": 2.721442885771543,
+      "grad_norm": 1.3507797718048096,
+      "learning_rate": 1.4107278040224886e-05,
+      "loss": 0.2297,
+      "step": 680
+    },
+    {
+      "epoch": 2.741482965931864,
+      "grad_norm": 1.330955147743225,
+      "learning_rate": 1.3909213888868244e-05,
+      "loss": 0.2366,
+      "step": 685
+    },
+    {
+      "epoch": 2.7615230460921842,
+      "grad_norm": 1.2785532474517822,
+      "learning_rate": 1.3711340754317088e-05,
+      "loss": 0.2169,
+      "step": 690
+    },
+    {
+      "epoch": 2.781563126252505,
+      "grad_norm": 1.2851275205612183,
+      "learning_rate": 1.3513693287813932e-05,
+      "loss": 0.2343,
+      "step": 695
+    },
+    {
+      "epoch": 2.8016032064128256,
+      "grad_norm": 1.5116199254989624,
+      "learning_rate": 1.3316306101082625e-05,
+      "loss": 0.2222,
+      "step": 700
+    },
+    {
+      "epoch": 2.8216432865731464,
+      "grad_norm": 1.1207387447357178,
+      "learning_rate": 1.3119213760267222e-05,
+      "loss": 0.2244,
+      "step": 705
+    },
+    {
+      "epoch": 2.841683366733467,
+      "grad_norm": 1.202231764793396,
+      "learning_rate": 1.2922450779878824e-05,
+      "loss": 0.238,
+      "step": 710
+    },
+    {
+      "epoch": 2.8617234468937873,
+      "grad_norm": 1.4184461832046509,
+      "learning_rate": 1.2726051616751415e-05,
+      "loss": 0.2129,
+      "step": 715
+    },
+    {
+      "epoch": 2.881763527054108,
+      "grad_norm": 1.2793796062469482,
+      "learning_rate": 1.253005066400786e-05,
+      "loss": 0.2107,
+      "step": 720
+    },
+    {
+      "epoch": 2.901803607214429,
+      "grad_norm": 1.4120863676071167,
+      "learning_rate": 1.2334482245037041e-05,
+      "loss": 0.2185,
+      "step": 725
+    },
+    {
+      "epoch": 2.9218436873747495,
+      "grad_norm": 1.1317527294158936,
+      "learning_rate": 1.2139380607483156e-05,
+      "loss": 0.2274,
+      "step": 730
+    },
+    {
+      "epoch": 2.94188376753507,
+      "grad_norm": 1.3882746696472168,
+      "learning_rate": 1.1944779917248379e-05,
+      "loss": 0.2226,
+      "step": 735
+    },
+    {
+      "epoch": 2.961923847695391,
+      "grad_norm": 1.3538494110107422,
+      "learning_rate": 1.1750714252509733e-05,
+      "loss": 0.2144,
+      "step": 740
+    },
+    {
+      "epoch": 2.9819639278557113,
+      "grad_norm": 1.2715141773223877,
+      "learning_rate": 1.1557217597751391e-05,
+      "loss": 0.1835,
+      "step": 745
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.9165490865707397,
+      "learning_rate": 1.1364323837813377e-05,
+      "loss": 0.2478,
+      "step": 750
+    },
+    {
+      "epoch": 3.0200400801603204,
+      "grad_norm": 1.1404417753219604,
+      "learning_rate": 1.1172066751957671e-05,
+      "loss": 0.1838,
+      "step": 755
+    },
+    {
+      "epoch": 3.0400801603206413,
+      "grad_norm": 1.239406943321228,
+      "learning_rate": 1.0980480007952866e-05,
+      "loss": 0.1478,
+      "step": 760
+    },
+    {
+      "epoch": 3.0601202404809618,
+      "grad_norm": 1.1695140600204468,
+      "learning_rate": 1.0789597156178318e-05,
+      "loss": 0.1623,
+      "step": 765
+    },
+    {
+      "epoch": 3.0801603206412826,
+      "grad_norm": 1.2266541719436646,
+      "learning_rate": 1.0599451623748823e-05,
+      "loss": 0.1702,
+      "step": 770
+    },
+    {
+      "epoch": 3.100200400801603,
+      "grad_norm": 1.154406189918518,
+      "learning_rate": 1.0410076708660959e-05,
+      "loss": 0.1585,
+      "step": 775
+    },
+    {
+      "epoch": 3.120240480961924,
+      "grad_norm": 1.3153400421142578,
+      "learning_rate": 1.0221505573961948e-05,
+      "loss": 0.146,
+      "step": 780
+    },
+    {
+      "epoch": 3.1402805611222444,
+      "grad_norm": 1.1898683309555054,
+      "learning_rate": 1.0033771241942221e-05,
+      "loss": 0.1324,
+      "step": 785
+    },
+    {
+      "epoch": 3.1603206412825653,
+      "grad_norm": 1.3161805868148804,
+      "learning_rate": 9.846906588352599e-06,
+      "loss": 0.1508,
+      "step": 790
+    },
+    {
+      "epoch": 3.1803607214428857,
+      "grad_norm": 1.435442328453064,
+      "learning_rate": 9.660944336647129e-06,
+      "loss": 0.1591,
+      "step": 795
+    },
+    {
+      "epoch": 3.2004008016032066,
+      "grad_norm": 1.415054202079773,
+      "learning_rate": 9.475917052252616e-06,
+      "loss": 0.158,
+      "step": 800
+    },
+    {
+      "epoch": 3.220440881763527,
+      "grad_norm": 1.1507920026779175,
+      "learning_rate": 9.291857136865814e-06,
+      "loss": 0.1514,
+      "step": 805
+    },
+    {
+      "epoch": 3.2404809619238475,
+      "grad_norm": 1.2796574831008911,
+      "learning_rate": 9.10879682277928e-06,
+      "loss": 0.1529,
+      "step": 810
+    },
+    {
+      "epoch": 3.2605210420841684,
+      "grad_norm": 1.2656141519546509,
+      "learning_rate": 8.926768167236924e-06,
+      "loss": 0.1402,
+      "step": 815
+    },
+    {
+      "epoch": 3.280561122244489,
+      "grad_norm": 1.0473889112472534,
+      "learning_rate": 8.745803046820168e-06,
+      "loss": 0.1807,
+      "step": 820
+    },
+    {
+      "epoch": 3.3006012024048097,
+      "grad_norm": 0.9760593771934509,
+      "learning_rate": 8.565933151865785e-06,
+      "loss": 0.1514,
+      "step": 825
+    },
+    {
+      "epoch": 3.32064128256513,
+      "grad_norm": 1.1415501832962036,
+      "learning_rate": 8.387189980916318e-06,
+      "loss": 0.1321,
+      "step": 830
+    },
+    {
+      "epoch": 3.340681362725451,
+      "grad_norm": 1.1905689239501953,
+      "learning_rate": 8.20960483520414e-06,
+      "loss": 0.1422,
+      "step": 835
+    },
+    {
+      "epoch": 3.3607214428857715,
+      "grad_norm": 1.0787945985794067,
+      "learning_rate": 8.033208813169986e-06,
+      "loss": 0.1378,
+      "step": 840
+    },
+    {
+      "epoch": 3.3807615230460923,
+      "grad_norm": 1.1386911869049072,
+      "learning_rate": 7.858032805017063e-06,
+      "loss": 0.1244,
+      "step": 845
+    },
+    {
+      "epoch": 3.400801603206413,
+      "grad_norm": 1.272212028503418,
+      "learning_rate": 7.684107487301606e-06,
+      "loss": 0.1422,
+      "step": 850
+    },
+    {
+      "epoch": 3.420841683366733,
+      "grad_norm": 1.29020094871521,
+      "learning_rate": 7.511463317560839e-06,
+      "loss": 0.1379,
+      "step": 855
+    },
+    {
+      "epoch": 3.440881763527054,
+      "grad_norm": 1.1315019130706787,
+      "learning_rate": 7.340130528979295e-06,
+      "loss": 0.1537,
+      "step": 860
+    },
+    {
+      "epoch": 3.460921843687375,
+      "grad_norm": 1.2238397598266602,
+      "learning_rate": 7.17013912509445e-06,
+      "loss": 0.1315,
+      "step": 865
+    },
+    {
+      "epoch": 3.4809619238476954,
+      "grad_norm": 1.290061593055725,
+      "learning_rate": 7.001518874542514e-06,
+      "loss": 0.162,
+      "step": 870
+    },
+    {
+      "epoch": 3.501002004008016,
+      "grad_norm": 1.1427054405212402,
+      "learning_rate": 6.834299305845414e-06,
+      "loss": 0.1381,
+      "step": 875
+    },
+    {
+      "epoch": 3.5210420841683367,
+      "grad_norm": 1.1416689157485962,
+      "learning_rate": 6.668509702239801e-06,
+      "loss": 0.1559,
+      "step": 880
+    },
+    {
+      "epoch": 3.541082164328657,
+      "grad_norm": 1.6092044115066528,
+      "learning_rate": 6.504179096549011e-06,
+      "loss": 0.1455,
+      "step": 885
+    },
+    {
+      "epoch": 3.561122244488978,
+      "grad_norm": 1.2860769033432007,
+      "learning_rate": 6.3413362660988934e-06,
+      "loss": 0.1379,
+      "step": 890
+    },
+    {
+      "epoch": 3.5811623246492985,
+      "grad_norm": 1.0260988473892212,
+      "learning_rate": 6.18000972767835e-06,
+      "loss": 0.1305,
+      "step": 895
+    },
+    {
+      "epoch": 3.6012024048096194,
+      "grad_norm": 1.1337883472442627,
+      "learning_rate": 6.0202277325455246e-06,
+      "loss": 0.1439,
+      "step": 900
+    },
+    {
+      "epoch": 3.62124248496994,
+      "grad_norm": 1.0394679307937622,
+      "learning_rate": 5.862018261480494e-06,
+      "loss": 0.1321,
+      "step": 905
+    },
+    {
+      "epoch": 3.6412825651302603,
+      "grad_norm": 1.0980677604675293,
+      "learning_rate": 5.7054090198853055e-06,
+      "loss": 0.1371,
+      "step": 910
+    },
+    {
+      "epoch": 3.661322645290581,
+      "grad_norm": 0.9747451543807983,
+      "learning_rate": 5.550427432932272e-06,
+      "loss": 0.1381,
+      "step": 915
+    },
+    {
+      "epoch": 3.681362725450902,
+      "grad_norm": 0.9877727627754211,
+      "learning_rate": 5.397100640761277e-06,
+      "loss": 0.1188,
+      "step": 920
+    },
+    {
+      "epoch": 3.7014028056112225,
+      "grad_norm": 1.2609772682189941,
+      "learning_rate": 5.24545549372707e-06,
+      "loss": 0.1424,
+      "step": 925
+    },
+    {
+      "epoch": 3.721442885771543,
+      "grad_norm": 1.0329722166061401,
+      "learning_rate": 5.0955185476972566e-06,
+      "loss": 0.124,
+      "step": 930
+    },
+    {
+      "epoch": 3.741482965931864,
+      "grad_norm": 1.0099831819534302,
+      "learning_rate": 4.9473160594018815e-06,
+      "loss": 0.1221,
+      "step": 935
+    },
+    {
+      "epoch": 3.7615230460921842,
+      "grad_norm": 1.1363469362258911,
+      "learning_rate": 4.8008739818353824e-06,
+      "loss": 0.1353,
+      "step": 940
+    },
+    {
+      "epoch": 3.781563126252505,
+      "grad_norm": 1.0320320129394531,
+      "learning_rate": 4.656217959711761e-06,
+      "loss": 0.1397,
+      "step": 945
+    },
+    {
+      "epoch": 3.8016032064128256,
+      "grad_norm": 1.3003877401351929,
+      "learning_rate": 4.513373324973691e-06,
+      "loss": 0.1284,
+      "step": 950
+    },
+    {
+      "epoch": 3.8216432865731464,
+      "grad_norm": 1.0216275453567505,
+      "learning_rate": 4.372365092356445e-06,
+      "loss": 0.1049,
+      "step": 955
+    },
+    {
+      "epoch": 3.841683366733467,
+      "grad_norm": 1.0927823781967163,
+      "learning_rate": 4.233217955007346e-06,
+      "loss": 0.1245,
+      "step": 960
+    },
+    {
+      "epoch": 3.8617234468937873,
+      "grad_norm": 1.0101195573806763,
+      "learning_rate": 4.0959562801615445e-06,
+      "loss": 0.1274,
+      "step": 965
+    },
+    {
+      "epoch": 3.881763527054108,
+      "grad_norm": 1.254036784172058,
+      "learning_rate": 3.960604104874871e-06,
+      "loss": 0.1392,
+      "step": 970
+    },
+    {
+      "epoch": 3.901803607214429,
+      "grad_norm": 1.62558114528656,
+      "learning_rate": 3.8271851318144865e-06,
+      "loss": 0.1318,
+      "step": 975
+    },
+    {
+      "epoch": 3.9218436873747495,
+      "grad_norm": 1.004096269607544,
+      "learning_rate": 3.695722725108121e-06,
+      "loss": 0.1221,
+      "step": 980
+    },
+    {
+      "epoch": 3.94188376753507,
+      "grad_norm": 1.1071776151657104,
+      "learning_rate": 3.566239906252576e-06,
+      "loss": 0.134,
+      "step": 985
+    },
+    {
+      "epoch": 3.961923847695391,
+      "grad_norm": 1.0152360200881958,
+      "learning_rate": 3.438759350082243e-06,
+      "loss": 0.1219,
+      "step": 990
+    },
+    {
+      "epoch": 3.9819639278557113,
+      "grad_norm": 1.0832067728042603,
+      "learning_rate": 3.31330338079831e-06,
+      "loss": 0.1198,
+      "step": 995
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.064515471458435,
+      "learning_rate": 3.1898939680593976e-06,
+      "loss": 0.1227,
+      "step": 1000
+    },
+    {
+      "epoch": 4.020040080160321,
+      "grad_norm": 0.9265746474266052,
+      "learning_rate": 3.0685527231342436e-06,
+      "loss": 0.0994,
+      "step": 1005
+    },
+    {
+      "epoch": 4.040080160320641,
+      "grad_norm": 0.8134159445762634,
+      "learning_rate": 2.9493008951171907e-06,
+      "loss": 0.1015,
+      "step": 1010
+    },
+    {
+      "epoch": 4.060120240480962,
+      "grad_norm": 0.8633864521980286,
+      "learning_rate": 2.8321593672070655e-06,
+      "loss": 0.1075,
+      "step": 1015
+    },
+    {
+      "epoch": 4.080160320641283,
+      "grad_norm": 0.8302706480026245,
+      "learning_rate": 2.7171486530501448e-06,
+      "loss": 0.0901,
+      "step": 1020
+    },
+    {
+      "epoch": 4.1002004008016035,
+      "grad_norm": 0.9754287004470825,
+      "learning_rate": 2.6042888931478505e-06,
+      "loss": 0.1021,
+      "step": 1025
+    },
+    {
+      "epoch": 4.1202404809619235,
+      "grad_norm": 0.8680105209350586,
+      "learning_rate": 2.493599851329758e-06,
+      "loss": 0.0859,
+      "step": 1030
+    },
+    {
+      "epoch": 4.140280561122244,
+      "grad_norm": 0.8552296161651611,
+      "learning_rate": 2.385100911292591e-06,
+      "loss": 0.09,
+      "step": 1035
+    },
+    {
+      "epoch": 4.160320641282565,
+      "grad_norm": 1.0095603466033936,
+      "learning_rate": 2.278811073205783e-06,
+      "loss": 0.1058,
+      "step": 1040
+    },
+    {
+      "epoch": 4.180360721442886,
+      "grad_norm": 0.9673905968666077,
+      "learning_rate": 2.174748950384186e-06,
+      "loss": 0.096,
+      "step": 1045
+    },
+    {
+      "epoch": 4.200400801603206,
+      "grad_norm": 0.9595734477043152,
+      "learning_rate": 2.0729327660285528e-06,
+      "loss": 0.0843,
+      "step": 1050
+    },
+    {
+      "epoch": 4.220440881763527,
+      "grad_norm": 0.8118200302124023,
+      "learning_rate": 1.9733803500343007e-06,
+      "loss": 0.0924,
+      "step": 1055
+    },
+    {
+      "epoch": 4.240480961923848,
+      "grad_norm": 0.8621242642402649,
+      "learning_rate": 1.8761091358691635e-06,
+      "loss": 0.0913,
+      "step": 1060
+    },
+    {
+      "epoch": 4.260521042084168,
+      "grad_norm": 0.9650718569755554,
+      "learning_rate": 1.7811361575202923e-06,
+      "loss": 0.0956,
+      "step": 1065
+    },
+    {
+      "epoch": 4.280561122244489,
+      "grad_norm": 1.0186007022857666,
+      "learning_rate": 1.6884780465112653e-06,
+      "loss": 0.1018,
+      "step": 1070
+    },
+    {
+      "epoch": 4.30060120240481,
+      "grad_norm": 0.8189525008201599,
+      "learning_rate": 1.5981510289896267e-06,
+      "loss": 0.0845,
+      "step": 1075
+    },
+    {
+      "epoch": 4.320641282565131,
+      "grad_norm": 1.0596710443496704,
+      "learning_rate": 1.510170922885357e-06,
+      "loss": 0.1027,
+      "step": 1080
+    },
+    {
+      "epoch": 4.340681362725451,
+      "grad_norm": 0.8731464743614197,
+      "learning_rate": 1.424553135140887e-06,
+      "loss": 0.0827,
+      "step": 1085
+    },
+    {
+      "epoch": 4.3607214428857715,
+      "grad_norm": 0.8803246021270752,
+      "learning_rate": 1.341312659013042e-06,
+      "loss": 0.0983,
+      "step": 1090
+    },
+    {
+      "epoch": 4.380761523046092,
+      "grad_norm": 0.7671328186988831,
+      "learning_rate": 1.2604640714474513e-06,
+      "loss": 0.0982,
+      "step": 1095
+    },
+    {
+      "epoch": 4.400801603206413,
+      "grad_norm": 0.7952969670295715,
+      "learning_rate": 1.1820215305258635e-06,
+      "loss": 0.091,
+      "step": 1100
+    },
+    {
+      "epoch": 4.420841683366733,
+      "grad_norm": 0.8785009384155273,
+      "learning_rate": 1.1059987729867888e-06,
+      "loss": 0.0801,
+      "step": 1105
+    },
+    {
+      "epoch": 4.440881763527054,
+      "grad_norm": 0.9525579214096069,
+      "learning_rate": 1.032409111819949e-06,
+      "loss": 0.1027,
+      "step": 1110
+    },
+    {
+      "epoch": 4.460921843687375,
+      "grad_norm": 0.7949389815330505,
+      "learning_rate": 9.612654339349358e-07,
+      "loss": 0.082,
+      "step": 1115
+    },
+    {
+      "epoch": 4.480961923847695,
+      "grad_norm": 0.9322835206985474,
+      "learning_rate": 8.925801979044573e-07,
+      "loss": 0.1009,
+      "step": 1120
+    },
+    {
+      "epoch": 4.501002004008016,
+      "grad_norm": 0.9823068976402283,
+      "learning_rate": 8.26365431782633e-07,
+      "loss": 0.0909,
+      "step": 1125
+    },
+    {
+      "epoch": 4.521042084168337,
+      "grad_norm": 0.7747249007225037,
+      "learning_rate": 7.626327309986414e-07,
+      "loss": 0.0847,
+      "step": 1130
+    },
+    {
+      "epoch": 4.541082164328658,
+      "grad_norm": 0.881176769733429,
+      "learning_rate": 7.013932563261494e-07,
+      "loss": 0.0928,
+      "step": 1135
+    },
+    {
+      "epoch": 4.561122244488978,
+      "grad_norm": 0.8490428328514099,
+      "learning_rate": 6.426577319288624e-07,
+      "loss": 0.0907,
+      "step": 1140
+    },
+    {
+      "epoch": 4.5811623246492985,
+      "grad_norm": 0.9037003517150879,
+      "learning_rate": 5.864364434825181e-07,
+      "loss": 0.1013,
+      "step": 1145
+    },
+    {
+      "epoch": 4.601202404809619,
+      "grad_norm": 0.9719511866569519,
+      "learning_rate": 5.327392363736777e-07,
+      "loss": 0.1032,
+      "step": 1150
+    },
+    {
+      "epoch": 4.62124248496994,
+      "grad_norm": 1.027663230895996,
+      "learning_rate": 4.815755139756195e-07,
+      "loss": 0.0858,
+      "step": 1155
+    },
+    {
+      "epoch": 4.64128256513026,
+      "grad_norm": 0.9042758941650391,
+      "learning_rate": 4.3295423600162475e-07,
+      "loss": 0.0825,
+      "step": 1160
+    },
+    {
+      "epoch": 4.661322645290581,
+      "grad_norm": 0.9892879724502563,
+      "learning_rate": 3.868839169359728e-07,
+      "loss": 0.1003,
+      "step": 1165
+    },
+    {
+      "epoch": 4.681362725450902,
+      "grad_norm": 0.8400679230690002,
+      "learning_rate": 3.43372624542902e-07,
+      "loss": 0.0928,
+      "step": 1170
+    },
+    {
+      "epoch": 4.701402805611222,
+      "grad_norm": 0.8155728578567505,
+      "learning_rate": 3.024279784537848e-07,
+      "loss": 0.0945,
+      "step": 1175
+    },
+    {
+      "epoch": 4.721442885771543,
+      "grad_norm": 0.8587477207183838,
+      "learning_rate": 2.640571488328003e-07,
+      "loss": 0.0933,
+      "step": 1180
+    },
+    {
+      "epoch": 4.741482965931864,
+      "grad_norm": 0.9381030201911926,
+      "learning_rate": 2.282668551212952e-07,
+      "loss": 0.0866,
+      "step": 1185
+    },
+    {
+      "epoch": 4.761523046092185,
+      "grad_norm": 0.8173649907112122,
+      "learning_rate": 1.95063364861095e-07,
+      "loss": 0.0857,
+      "step": 1190
+    },
+    {
+      "epoch": 4.781563126252505,
+      "grad_norm": 0.8224903345108032,
+      "learning_rate": 1.6445249259693374e-07,
+      "loss": 0.0893,
+      "step": 1195
+    },
+    {
+      "epoch": 4.801603206412826,
+      "grad_norm": 0.9319766759872437,
+      "learning_rate": 1.3643959885822677e-07,
+      "loss": 0.0945,
+      "step": 1200
+    },
+    {
+      "epoch": 4.8216432865731464,
+      "grad_norm": 1.0199525356292725,
+      "learning_rate": 1.1102958922034e-07,
+      "loss": 0.0932,
+      "step": 1205
+    },
+    {
+      "epoch": 4.841683366733466,
+      "grad_norm": 0.9463343024253845,
+      "learning_rate": 8.82269134455288e-08,
+      "loss": 0.0909,
+      "step": 1210
+    },
+    {
+      "epoch": 4.861723446893787,
+      "grad_norm": 0.88401859998703,
+      "learning_rate": 6.803556470370953e-08,
+      "loss": 0.0896,
+      "step": 1215
+    },
+    {
+      "epoch": 4.881763527054108,
+      "grad_norm": 0.9881237745285034,
+      "learning_rate": 5.0459078873177354e-08,
+      "loss": 0.1003,
+      "step": 1220
+    },
+    {
+      "epoch": 4.901803607214429,
+      "grad_norm": 0.906802237033844,
+      "learning_rate": 3.550053392140984e-08,
+      "loss": 0.0877,
+      "step": 1225
+    },
+    {
+      "epoch": 4.92184368737475,
+      "grad_norm": 1.0449682474136353,
+      "learning_rate": 2.3162549366058105e-08,
+      "loss": 0.0883,
+      "step": 1230
+    },
+    {
+      "epoch": 4.94188376753507,
+      "grad_norm": 0.9145968556404114,
+      "learning_rate": 1.3447285816222122e-08,
+      "loss": 0.0956,
+      "step": 1235
+    },
+    {
+      "epoch": 4.961923847695391,
+      "grad_norm": 0.861686110496521,
+      "learning_rate": 6.356444594081689e-09,
+      "loss": 0.0916,
+      "step": 1240
+    },
+    {
+      "epoch": 4.981963927855712,
+      "grad_norm": 0.9143417477607727,
+      "learning_rate": 1.891267436973099e-09,
+      "loss": 0.0981,
+      "step": 1245
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 1.2572712898254395,
+      "learning_rate": 5.2536279931403306e-11,
+      "loss": 0.0929,
+      "step": 1250
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1250,
+  "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": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6709372385207255e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

50_128_e5_3e-5/checkpoint-1250/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61c16f5cfa671126a0a7b100dbd969a49a07ef964c5bc306796be064502b8d67
+size 7736

50_128_e5_3e-5/checkpoint-1250/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)

50_128_e5_3e-5/checkpoint-250/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

50_128_e5_3e-5/checkpoint-250/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": [
+    "o_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

50_128_e5_3e-5/checkpoint-250/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61cc9afe09ef6287f220d65baf48654f434dd986bb9daca68d6bb6c808fd2618
+size 791751704

50_128_e5_3e-5/checkpoint-250/latest

@@ -0,0 +1 @@
+global_step250

50_128_e5_3e-5/checkpoint-250/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1651b80c28f388058d3979ba8fb8b9da7a3d0e21c9db347b5412ac1cebdf9804
+size 15920

50_128_e5_3e-5/checkpoint-250/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13fe36685e16b73417a283f8c7659e1b394e41aeb9141015fab8184bb5b35c05
+size 15920

50_128_e5_3e-5/checkpoint-250/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e38250e4d99e47a22fa7b376082c59791bebed94ba716585f7846707e5f74826
+size 15920