Upload folder using huggingface_hub

26_128_e5_3e-5/checkpoint-1090/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

26_128_e5_3e-5/checkpoint-1090/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

26_128_e5_3e-5/checkpoint-1090/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09daf30b9df030e4ff826cd8c9dafd552535377698954a5e8caff173631f05c5
+size 791751704

26_128_e5_3e-5/checkpoint-1090/latest

@@ -0,0 +1 @@
+global_step1090

26_128_e5_3e-5/checkpoint-1090/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:087f03960fc04e93c09f8119119e8291f04844955fd234f3031bc82841da3469
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f673fb966944c72f83877fca24d719053df0bec902f803c8bd2fd2116b2d15a
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b865bb48f8cfade09121e0a1b9160e6d29a3ebcff327ba4162fda050c51f4a
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cb4d79263fe48733d5164d26f075f42099e64c2a4ade727e307c5ecdda5b37b
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea6dae681749a61de3137eb8c8e6cd20d9e42257c027d1f97e9d9a5de95b0d7b
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:accff702631dc02874e46ae0cc44492519f8a12b34f3ddba382e31852e0804a2
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ead222da4e6cfe6bc9e480c9cccf0f3e30aa1263ea99e479c5dbfa95c8e21d9
+size 15920

26_128_e5_3e-5/checkpoint-1090/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0440e29373989c47f9963ca31ea82826e0412c2dbcaebf254e4b4b3fc52c9b4f
+size 15920

26_128_e5_3e-5/checkpoint-1090/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c88f4adb462425e9a93734b2fd7b35ba0a5d05d2e43afc9e9a19284ab8b91af7
+size 1064

26_128_e5_3e-5/checkpoint-1090/special_tokens_map.json

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

26_128_e5_3e-5/checkpoint-1090/tokenizer_config.json

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

26_128_e5_3e-5/checkpoint-1090/trainer_state.json

@@ -0,0 +1,1560 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 1090,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.022935779816513763,
+      "grad_norm": 1.1547188758850098,
+      "learning_rate": 2.181818181818182e-06,
+      "loss": 1.3697,
+      "step": 5
+    },
+    {
+      "epoch": 0.045871559633027525,
+      "grad_norm": 0.7839906811714172,
+      "learning_rate": 4.90909090909091e-06,
+      "loss": 1.3117,
+      "step": 10
+    },
+    {
+      "epoch": 0.06880733944954129,
+      "grad_norm": 0.7896807193756104,
+      "learning_rate": 7.636363636363636e-06,
+      "loss": 1.3411,
+      "step": 15
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.4968399405479431,
+      "learning_rate": 1.0363636363636364e-05,
+      "loss": 1.3191,
+      "step": 20
+    },
+    {
+      "epoch": 0.11467889908256881,
+      "grad_norm": 0.48551270365715027,
+      "learning_rate": 1.309090909090909e-05,
+      "loss": 1.2715,
+      "step": 25
+    },
+    {
+      "epoch": 0.13761467889908258,
+      "grad_norm": 0.5304503440856934,
+      "learning_rate": 1.5818181818181818e-05,
+      "loss": 1.2378,
+      "step": 30
+    },
+    {
+      "epoch": 0.16055045871559634,
+      "grad_norm": 0.5629490613937378,
+      "learning_rate": 1.8545454545454545e-05,
+      "loss": 1.2796,
+      "step": 35
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.5019248723983765,
+      "learning_rate": 2.1272727272727273e-05,
+      "loss": 1.2179,
+      "step": 40
+    },
+    {
+      "epoch": 0.20642201834862386,
+      "grad_norm": 0.566190779209137,
+      "learning_rate": 2.4e-05,
+      "loss": 1.2287,
+      "step": 45
+    },
+    {
+      "epoch": 0.22935779816513763,
+      "grad_norm": 0.5070985555648804,
+      "learning_rate": 2.6727272727272728e-05,
+      "loss": 1.1711,
+      "step": 50
+    },
+    {
+      "epoch": 0.25229357798165136,
+      "grad_norm": 0.482239305973053,
+      "learning_rate": 2.9454545454545456e-05,
+      "loss": 1.127,
+      "step": 55
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.5259355306625366,
+      "learning_rate": 2.999889440782374e-05,
+      "loss": 1.1649,
+      "step": 60
+    },
+    {
+      "epoch": 0.2981651376146789,
+      "grad_norm": 0.48876306414604187,
+      "learning_rate": 2.9994403218928662e-05,
+      "loss": 1.1244,
+      "step": 65
+    },
+    {
+      "epoch": 0.3211009174311927,
+      "grad_norm": 0.5676949620246887,
+      "learning_rate": 2.9986458367463977e-05,
+      "loss": 1.0935,
+      "step": 70
+    },
+    {
+      "epoch": 0.3440366972477064,
+      "grad_norm": 0.5986928939819336,
+      "learning_rate": 2.997506168336822e-05,
+      "loss": 1.0729,
+      "step": 75
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.6229660511016846,
+      "learning_rate": 2.9960215791640898e-05,
+      "loss": 1.0235,
+      "step": 80
+    },
+    {
+      "epoch": 0.38990825688073394,
+      "grad_norm": 0.602012574672699,
+      "learning_rate": 2.9941924111737924e-05,
+      "loss": 0.9924,
+      "step": 85
+    },
+    {
+      "epoch": 0.41284403669724773,
+      "grad_norm": 0.6534562706947327,
+      "learning_rate": 2.9920190856783978e-05,
+      "loss": 1.0184,
+      "step": 90
+    },
+    {
+      "epoch": 0.43577981651376146,
+      "grad_norm": 0.6122053265571594,
+      "learning_rate": 2.98950210326021e-05,
+      "loss": 1.0002,
+      "step": 95
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.7329753041267395,
+      "learning_rate": 2.9866420436560697e-05,
+      "loss": 0.9922,
+      "step": 100
+    },
+    {
+      "epoch": 0.481651376146789,
+      "grad_norm": 0.7041805386543274,
+      "learning_rate": 2.9834395656238264e-05,
+      "loss": 0.9824,
+      "step": 105
+    },
+    {
+      "epoch": 0.5045871559633027,
+      "grad_norm": 0.6798579692840576,
+      "learning_rate": 2.979895406790603e-05,
+      "loss": 0.9195,
+      "step": 110
+    },
+    {
+      "epoch": 0.5275229357798165,
+      "grad_norm": 0.6428303122520447,
+      "learning_rate": 2.9760103834828996e-05,
+      "loss": 0.9656,
+      "step": 115
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.7067289352416992,
+      "learning_rate": 2.971785390538569e-05,
+      "loss": 0.8735,
+      "step": 120
+    },
+    {
+      "epoch": 0.573394495412844,
+      "grad_norm": 0.9809579253196716,
+      "learning_rate": 2.9672214011007087e-05,
+      "loss": 0.8944,
+      "step": 125
+    },
+    {
+      "epoch": 0.5963302752293578,
+      "grad_norm": 0.7807005643844604,
+      "learning_rate": 2.9623194663935153e-05,
+      "loss": 0.8463,
+      "step": 130
+    },
+    {
+      "epoch": 0.6192660550458715,
+      "grad_norm": 0.8913285136222839,
+      "learning_rate": 2.9570807154801587e-05,
+      "loss": 0.8572,
+      "step": 135
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.9661921262741089,
+      "learning_rate": 2.951506355002722e-05,
+      "loss": 0.8317,
+      "step": 140
+    },
+    {
+      "epoch": 0.6651376146788991,
+      "grad_norm": 0.785257875919342,
+      "learning_rate": 2.9455976689042774e-05,
+      "loss": 0.7877,
+      "step": 145
+    },
+    {
+      "epoch": 0.6880733944954128,
+      "grad_norm": 0.8387299180030823,
+      "learning_rate": 2.9393560181331543e-05,
+      "loss": 0.7675,
+      "step": 150
+    },
+    {
+      "epoch": 0.7110091743119266,
+      "grad_norm": 0.83286452293396,
+      "learning_rate": 2.9327828403294725e-05,
+      "loss": 0.767,
+      "step": 155
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.97318035364151,
+      "learning_rate": 2.92587964949401e-05,
+      "loss": 0.7821,
+      "step": 160
+    },
+    {
+      "epoch": 0.7568807339449541,
+      "grad_norm": 1.0256941318511963,
+      "learning_rate": 2.918648035639484e-05,
+      "loss": 0.7215,
+      "step": 165
+    },
+    {
+      "epoch": 0.7798165137614679,
+      "grad_norm": 0.8714763522148132,
+      "learning_rate": 2.9110896644243226e-05,
+      "loss": 0.7473,
+      "step": 170
+    },
+    {
+      "epoch": 0.8027522935779816,
+      "grad_norm": 0.914694607257843,
+      "learning_rate": 2.9032062767690116e-05,
+      "loss": 0.7344,
+      "step": 175
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 1.7585707902908325,
+      "learning_rate": 2.8949996884551118e-05,
+      "loss": 0.676,
+      "step": 180
+    },
+    {
+      "epoch": 0.8486238532110092,
+      "grad_norm": 0.9376805424690247,
+      "learning_rate": 2.886471789707026e-05,
+      "loss": 0.7359,
+      "step": 185
+    },
+    {
+      "epoch": 0.8715596330275229,
+      "grad_norm": 0.948411226272583,
+      "learning_rate": 2.8776245447566274e-05,
+      "loss": 0.6896,
+      "step": 190
+    },
+    {
+      "epoch": 0.8944954128440367,
+      "grad_norm": 1.0342053174972534,
+      "learning_rate": 2.868459991390833e-05,
+      "loss": 0.688,
+      "step": 195
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 1.0501846075057983,
+      "learning_rate": 2.8589802404822457e-05,
+      "loss": 0.6407,
+      "step": 200
+    },
+    {
+      "epoch": 0.9403669724770642,
+      "grad_norm": 0.932583212852478,
+      "learning_rate": 2.8491874755029502e-05,
+      "loss": 0.6745,
+      "step": 205
+    },
+    {
+      "epoch": 0.963302752293578,
+      "grad_norm": 1.0710359811782837,
+      "learning_rate": 2.8390839520216004e-05,
+      "loss": 0.6641,
+      "step": 210
+    },
+    {
+      "epoch": 0.9862385321100917,
+      "grad_norm": 1.0051730871200562,
+      "learning_rate": 2.8286719971838885e-05,
+      "loss": 0.6803,
+      "step": 215
+    },
+    {
+      "epoch": 1.0091743119266054,
+      "grad_norm": 1.0668774843215942,
+      "learning_rate": 2.817954009176536e-05,
+      "loss": 0.6208,
+      "step": 220
+    },
+    {
+      "epoch": 1.0321100917431192,
+      "grad_norm": 1.1521594524383545,
+      "learning_rate": 2.806932456674919e-05,
+      "loss": 0.5436,
+      "step": 225
+    },
+    {
+      "epoch": 1.0550458715596331,
+      "grad_norm": 1.4106714725494385,
+      "learning_rate": 2.7956098782744572e-05,
+      "loss": 0.5521,
+      "step": 230
+    },
+    {
+      "epoch": 1.0779816513761469,
+      "grad_norm": 1.2204782962799072,
+      "learning_rate": 2.7839888819058998e-05,
+      "loss": 0.5417,
+      "step": 235
+    },
+    {
+      "epoch": 1.1009174311926606,
+      "grad_norm": 1.1560258865356445,
+      "learning_rate": 2.772072144234639e-05,
+      "loss": 0.5413,
+      "step": 240
+    },
+    {
+      "epoch": 1.1238532110091743,
+      "grad_norm": 1.0572724342346191,
+      "learning_rate": 2.759862410044195e-05,
+      "loss": 0.5248,
+      "step": 245
+    },
+    {
+      "epoch": 1.146788990825688,
+      "grad_norm": 1.0982731580734253,
+      "learning_rate": 2.747362491604008e-05,
+      "loss": 0.5269,
+      "step": 250
+    },
+    {
+      "epoch": 1.1697247706422018,
+      "grad_norm": 1.179128885269165,
+      "learning_rate": 2.734575268021691e-05,
+      "loss": 0.5163,
+      "step": 255
+    },
+    {
+      "epoch": 1.1926605504587156,
+      "grad_norm": 1.1007860898971558,
+      "learning_rate": 2.72150368457988e-05,
+      "loss": 0.4929,
+      "step": 260
+    },
+    {
+      "epoch": 1.2155963302752293,
+      "grad_norm": 1.113296627998352,
+      "learning_rate": 2.708150752057851e-05,
+      "loss": 0.5457,
+      "step": 265
+    },
+    {
+      "epoch": 1.238532110091743,
+      "grad_norm": 1.220629334449768,
+      "learning_rate": 2.6945195460380446e-05,
+      "loss": 0.4796,
+      "step": 270
+    },
+    {
+      "epoch": 1.261467889908257,
+      "grad_norm": 1.1990607976913452,
+      "learning_rate": 2.6806132061976664e-05,
+      "loss": 0.4461,
+      "step": 275
+    },
+    {
+      "epoch": 1.2844036697247707,
+      "grad_norm": 1.211341142654419,
+      "learning_rate": 2.6664349355855267e-05,
+      "loss": 0.5242,
+      "step": 280
+    },
+    {
+      "epoch": 1.3073394495412844,
+      "grad_norm": 1.222888708114624,
+      "learning_rate": 2.6519879998842813e-05,
+      "loss": 0.4824,
+      "step": 285
+    },
+    {
+      "epoch": 1.3302752293577982,
+      "grad_norm": 1.3487645387649536,
+      "learning_rate": 2.6372757266582443e-05,
+      "loss": 0.5034,
+      "step": 290
+    },
+    {
+      "epoch": 1.353211009174312,
+      "grad_norm": 1.1871864795684814,
+      "learning_rate": 2.6223015045869528e-05,
+      "loss": 0.478,
+      "step": 295
+    },
+    {
+      "epoch": 1.3761467889908257,
+      "grad_norm": 1.238922357559204,
+      "learning_rate": 2.6070687826846508e-05,
+      "loss": 0.4627,
+      "step": 300
+    },
+    {
+      "epoch": 1.3990825688073394,
+      "grad_norm": 1.1637448072433472,
+      "learning_rate": 2.591581069505876e-05,
+      "loss": 0.4875,
+      "step": 305
+    },
+    {
+      "epoch": 1.4220183486238533,
+      "grad_norm": 1.4148343801498413,
+      "learning_rate": 2.5758419323373373e-05,
+      "loss": 0.4189,
+      "step": 310
+    },
+    {
+      "epoch": 1.4449541284403669,
+      "grad_norm": 1.0583734512329102,
+      "learning_rate": 2.5598549963762578e-05,
+      "loss": 0.4594,
+      "step": 315
+    },
+    {
+      "epoch": 1.4678899082568808,
+      "grad_norm": 1.0729453563690186,
+      "learning_rate": 2.5436239438953838e-05,
+      "loss": 0.4897,
+      "step": 320
+    },
+    {
+      "epoch": 1.4908256880733946,
+      "grad_norm": 1.1639119386672974,
+      "learning_rate": 2.5271525133948484e-05,
+      "loss": 0.4402,
+      "step": 325
+    },
+    {
+      "epoch": 1.5137614678899083,
+      "grad_norm": 1.1435127258300781,
+      "learning_rate": 2.5104444987410803e-05,
+      "loss": 0.453,
+      "step": 330
+    },
+    {
+      "epoch": 1.536697247706422,
+      "grad_norm": 1.2128199338912964,
+      "learning_rate": 2.4935037482929616e-05,
+      "loss": 0.4432,
+      "step": 335
+    },
+    {
+      "epoch": 1.5596330275229358,
+      "grad_norm": 1.266803503036499,
+      "learning_rate": 2.4763341640154375e-05,
+      "loss": 0.3958,
+      "step": 340
+    },
+    {
+      "epoch": 1.5825688073394495,
+      "grad_norm": 1.2251777648925781,
+      "learning_rate": 2.458939700580774e-05,
+      "loss": 0.4251,
+      "step": 345
+    },
+    {
+      "epoch": 1.6055045871559632,
+      "grad_norm": 1.2751761674880981,
+      "learning_rate": 2.441324364457678e-05,
+      "loss": 0.3917,
+      "step": 350
+    },
+    {
+      "epoch": 1.6284403669724772,
+      "grad_norm": 1.1549896001815796,
+      "learning_rate": 2.4234922129884873e-05,
+      "loss": 0.4136,
+      "step": 355
+    },
+    {
+      "epoch": 1.6513761467889907,
+      "grad_norm": 1.1309773921966553,
+      "learning_rate": 2.405447353454645e-05,
+      "loss": 0.4608,
+      "step": 360
+    },
+    {
+      "epoch": 1.6743119266055047,
+      "grad_norm": 1.081762433052063,
+      "learning_rate": 2.3871939421306633e-05,
+      "loss": 0.3717,
+      "step": 365
+    },
+    {
+      "epoch": 1.6972477064220184,
+      "grad_norm": 1.3630900382995605,
+      "learning_rate": 2.368736183326814e-05,
+      "loss": 0.3734,
+      "step": 370
+    },
+    {
+      "epoch": 1.7201834862385321,
+      "grad_norm": 1.2541098594665527,
+      "learning_rate": 2.3500783284207445e-05,
+      "loss": 0.3897,
+      "step": 375
+    },
+    {
+      "epoch": 1.7431192660550459,
+      "grad_norm": 1.2859941720962524,
+      "learning_rate": 2.3312246748782624e-05,
+      "loss": 0.3775,
+      "step": 380
+    },
+    {
+      "epoch": 1.7660550458715596,
+      "grad_norm": 1.1434638500213623,
+      "learning_rate": 2.3121795652634945e-05,
+      "loss": 0.3652,
+      "step": 385
+    },
+    {
+      "epoch": 1.7889908256880735,
+      "grad_norm": 1.16877019405365,
+      "learning_rate": 2.2929473862386684e-05,
+      "loss": 0.3491,
+      "step": 390
+    },
+    {
+      "epoch": 1.811926605504587,
+      "grad_norm": 1.2541364431381226,
+      "learning_rate": 2.273532567553728e-05,
+      "loss": 0.3713,
+      "step": 395
+    },
+    {
+      "epoch": 1.834862385321101,
+      "grad_norm": 1.2320799827575684,
+      "learning_rate": 2.2539395810260316e-05,
+      "loss": 0.3513,
+      "step": 400
+    },
+    {
+      "epoch": 1.8577981651376145,
+      "grad_norm": 1.2168241739273071,
+      "learning_rate": 2.234172939510356e-05,
+      "loss": 0.35,
+      "step": 405
+    },
+    {
+      "epoch": 1.8807339449541285,
+      "grad_norm": 1.3372358083724976,
+      "learning_rate": 2.2142371958594526e-05,
+      "loss": 0.3656,
+      "step": 410
+    },
+    {
+      "epoch": 1.9036697247706422,
+      "grad_norm": 1.2136768102645874,
+      "learning_rate": 2.1941369418753836e-05,
+      "loss": 0.3534,
+      "step": 415
+    },
+    {
+      "epoch": 1.926605504587156,
+      "grad_norm": 1.1403114795684814,
+      "learning_rate": 2.1738768072518965e-05,
+      "loss": 0.3589,
+      "step": 420
+    },
+    {
+      "epoch": 1.9495412844036697,
+      "grad_norm": 1.1859922409057617,
+      "learning_rate": 2.1534614585080636e-05,
+      "loss": 0.3481,
+      "step": 425
+    },
+    {
+      "epoch": 1.9724770642201834,
+      "grad_norm": 1.1227407455444336,
+      "learning_rate": 2.1328955979134443e-05,
+      "loss": 0.3189,
+      "step": 430
+    },
+    {
+      "epoch": 1.9954128440366974,
+      "grad_norm": 1.181204080581665,
+      "learning_rate": 2.112183962405008e-05,
+      "loss": 0.303,
+      "step": 435
+    },
+    {
+      "epoch": 2.018348623853211,
+      "grad_norm": 1.1871689558029175,
+      "learning_rate": 2.0913313224960757e-05,
+      "loss": 0.2873,
+      "step": 440
+    },
+    {
+      "epoch": 2.041284403669725,
+      "grad_norm": 1.223633885383606,
+      "learning_rate": 2.07034248117753e-05,
+      "loss": 0.2558,
+      "step": 445
+    },
+    {
+      "epoch": 2.0642201834862384,
+      "grad_norm": 1.118867039680481,
+      "learning_rate": 2.0492222728115373e-05,
+      "loss": 0.2364,
+      "step": 450
+    },
+    {
+      "epoch": 2.0871559633027523,
+      "grad_norm": 1.2636170387268066,
+      "learning_rate": 2.0279755620180518e-05,
+      "loss": 0.2657,
+      "step": 455
+    },
+    {
+      "epoch": 2.1100917431192663,
+      "grad_norm": 1.1175693273544312,
+      "learning_rate": 2.0066072425543455e-05,
+      "loss": 0.2527,
+      "step": 460
+    },
+    {
+      "epoch": 2.13302752293578,
+      "grad_norm": 1.0788837671279907,
+      "learning_rate": 1.985122236187829e-05,
+      "loss": 0.2559,
+      "step": 465
+    },
+    {
+      "epoch": 2.1559633027522938,
+      "grad_norm": 1.2557427883148193,
+      "learning_rate": 1.963525491562421e-05,
+      "loss": 0.2788,
+      "step": 470
+    },
+    {
+      "epoch": 2.1788990825688073,
+      "grad_norm": 1.2082065343856812,
+      "learning_rate": 1.9418219830587252e-05,
+      "loss": 0.2722,
+      "step": 475
+    },
+    {
+      "epoch": 2.2018348623853212,
+      "grad_norm": 1.1580853462219238,
+      "learning_rate": 1.9200167096482812e-05,
+      "loss": 0.2691,
+      "step": 480
+    },
+    {
+      "epoch": 2.2247706422018347,
+      "grad_norm": 1.301430106163025,
+      "learning_rate": 1.8981146937421516e-05,
+      "loss": 0.2483,
+      "step": 485
+    },
+    {
+      "epoch": 2.2477064220183487,
+      "grad_norm": 1.255772590637207,
+      "learning_rate": 1.8761209800341076e-05,
+      "loss": 0.2555,
+      "step": 490
+    },
+    {
+      "epoch": 2.270642201834862,
+      "grad_norm": 1.2370141744613647,
+      "learning_rate": 1.8540406343386865e-05,
+      "loss": 0.2804,
+      "step": 495
+    },
+    {
+      "epoch": 2.293577981651376,
+      "grad_norm": 1.165006399154663,
+      "learning_rate": 1.8318787424243817e-05,
+      "loss": 0.2043,
+      "step": 500
+    },
+    {
+      "epoch": 2.31651376146789,
+      "grad_norm": 1.171410322189331,
+      "learning_rate": 1.8096404088422335e-05,
+      "loss": 0.2691,
+      "step": 505
+    },
+    {
+      "epoch": 2.3394495412844036,
+      "grad_norm": 1.136547565460205,
+      "learning_rate": 1.7873307557501033e-05,
+      "loss": 0.2289,
+      "step": 510
+    },
+    {
+      "epoch": 2.3623853211009176,
+      "grad_norm": 1.1734422445297241,
+      "learning_rate": 1.7649549217328815e-05,
+      "loss": 0.2124,
+      "step": 515
+    },
+    {
+      "epoch": 2.385321100917431,
+      "grad_norm": 1.1301604509353638,
+      "learning_rate": 1.7425180606189174e-05,
+      "loss": 0.2335,
+      "step": 520
+    },
+    {
+      "epoch": 2.408256880733945,
+      "grad_norm": 1.1791046857833862,
+      "learning_rate": 1.7200253402929412e-05,
+      "loss": 0.2068,
+      "step": 525
+    },
+    {
+      "epoch": 2.4311926605504586,
+      "grad_norm": 1.1526075601577759,
+      "learning_rate": 1.6974819415057387e-05,
+      "loss": 0.2133,
+      "step": 530
+    },
+    {
+      "epoch": 2.4541284403669725,
+      "grad_norm": 1.1090424060821533,
+      "learning_rate": 1.6748930566808733e-05,
+      "loss": 0.2043,
+      "step": 535
+    },
+    {
+      "epoch": 2.477064220183486,
+      "grad_norm": 1.287276268005371,
+      "learning_rate": 1.652263888718712e-05,
+      "loss": 0.2156,
+      "step": 540
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.3875986337661743,
+      "learning_rate": 1.6295996497980407e-05,
+      "loss": 0.1913,
+      "step": 545
+    },
+    {
+      "epoch": 2.522935779816514,
+      "grad_norm": 1.1331018209457397,
+      "learning_rate": 1.606905560175544e-05,
+      "loss": 0.2229,
+      "step": 550
+    },
+    {
+      "epoch": 2.5458715596330275,
+      "grad_norm": 1.1809563636779785,
+      "learning_rate": 1.584186846983421e-05,
+      "loss": 0.2174,
+      "step": 555
+    },
+    {
+      "epoch": 2.5688073394495414,
+      "grad_norm": 1.1611757278442383,
+      "learning_rate": 1.5614487430254216e-05,
+      "loss": 0.1976,
+      "step": 560
+    },
+    {
+      "epoch": 2.591743119266055,
+      "grad_norm": 1.21211576461792,
+      "learning_rate": 1.5386964855715714e-05,
+      "loss": 0.2291,
+      "step": 565
+    },
+    {
+      "epoch": 2.614678899082569,
+      "grad_norm": 1.3323678970336914,
+      "learning_rate": 1.5159353151518745e-05,
+      "loss": 0.2303,
+      "step": 570
+    },
+    {
+      "epoch": 2.6376146788990824,
+      "grad_norm": 1.1840834617614746,
+      "learning_rate": 1.4931704743492567e-05,
+      "loss": 0.1748,
+      "step": 575
+    },
+    {
+      "epoch": 2.6605504587155964,
+      "grad_norm": 1.1386724710464478,
+      "learning_rate": 1.4704072065920458e-05,
+      "loss": 0.1843,
+      "step": 580
+    },
+    {
+      "epoch": 2.68348623853211,
+      "grad_norm": 1.1680500507354736,
+      "learning_rate": 1.447650754946249e-05,
+      "loss": 0.182,
+      "step": 585
+    },
+    {
+      "epoch": 2.706422018348624,
+      "grad_norm": 1.0924367904663086,
+      "learning_rate": 1.4249063609079167e-05,
+      "loss": 0.1956,
+      "step": 590
+    },
+    {
+      "epoch": 2.729357798165138,
+      "grad_norm": 1.127066731452942,
+      "learning_rate": 1.4021792631958714e-05,
+      "loss": 0.2049,
+      "step": 595
+    },
+    {
+      "epoch": 2.7522935779816513,
+      "grad_norm": 1.1841648817062378,
+      "learning_rate": 1.379474696545069e-05,
+      "loss": 0.187,
+      "step": 600
+    },
+    {
+      "epoch": 2.7752293577981653,
+      "grad_norm": 1.1207882165908813,
+      "learning_rate": 1.3567978905008849e-05,
+      "loss": 0.1834,
+      "step": 605
+    },
+    {
+      "epoch": 2.7981651376146788,
+      "grad_norm": 1.149422287940979,
+      "learning_rate": 1.3341540682145904e-05,
+      "loss": 0.2271,
+      "step": 610
+    },
+    {
+      "epoch": 2.8211009174311927,
+      "grad_norm": 0.9951650500297546,
+      "learning_rate": 1.3115484452403026e-05,
+      "loss": 0.1806,
+      "step": 615
+    },
+    {
+      "epoch": 2.8440366972477067,
+      "grad_norm": 1.1563984155654907,
+      "learning_rate": 1.2889862283336894e-05,
+      "loss": 0.1849,
+      "step": 620
+    },
+    {
+      "epoch": 2.86697247706422,
+      "grad_norm": 1.0431939363479614,
+      "learning_rate": 1.2664726142526931e-05,
+      "loss": 0.1606,
+      "step": 625
+    },
+    {
+      "epoch": 2.8899082568807337,
+      "grad_norm": 1.065514326095581,
+      "learning_rate": 1.2440127885605612e-05,
+      "loss": 0.1858,
+      "step": 630
+    },
+    {
+      "epoch": 2.9128440366972477,
+      "grad_norm": 1.2608879804611206,
+      "learning_rate": 1.2216119244314567e-05,
+      "loss": 0.1686,
+      "step": 635
+    },
+    {
+      "epoch": 2.9357798165137616,
+      "grad_norm": 1.1640936136245728,
+      "learning_rate": 1.1992751814589179e-05,
+      "loss": 0.1775,
+      "step": 640
+    },
+    {
+      "epoch": 2.958715596330275,
+      "grad_norm": 1.1765121221542358,
+      "learning_rate": 1.1770077044674508e-05,
+      "loss": 0.2141,
+      "step": 645
+    },
+    {
+      "epoch": 2.981651376146789,
+      "grad_norm": 1.033565640449524,
+      "learning_rate": 1.1548146223275206e-05,
+      "loss": 0.1934,
+      "step": 650
+    },
+    {
+      "epoch": 3.0045871559633026,
+      "grad_norm": 1.0922577381134033,
+      "learning_rate": 1.1327010467742167e-05,
+      "loss": 0.1838,
+      "step": 655
+    },
+    {
+      "epoch": 3.0275229357798166,
+      "grad_norm": 1.1589926481246948,
+      "learning_rate": 1.1106720712298694e-05,
+      "loss": 0.1486,
+      "step": 660
+    },
+    {
+      "epoch": 3.05045871559633,
+      "grad_norm": 1.0292847156524658,
+      "learning_rate": 1.0887327696308773e-05,
+      "loss": 0.1428,
+      "step": 665
+    },
+    {
+      "epoch": 3.073394495412844,
+      "grad_norm": 0.8822467923164368,
+      "learning_rate": 1.0668881952590303e-05,
+      "loss": 0.1087,
+      "step": 670
+    },
+    {
+      "epoch": 3.096330275229358,
+      "grad_norm": 1.0221686363220215,
+      "learning_rate": 1.0451433795775863e-05,
+      "loss": 0.1256,
+      "step": 675
+    },
+    {
+      "epoch": 3.1192660550458715,
+      "grad_norm": 0.9597510695457458,
+      "learning_rate": 1.0235033310723711e-05,
+      "loss": 0.1168,
+      "step": 680
+    },
+    {
+      "epoch": 3.1422018348623855,
+      "grad_norm": 1.1529306173324585,
+      "learning_rate": 1.0019730340981772e-05,
+      "loss": 0.1282,
+      "step": 685
+    },
+    {
+      "epoch": 3.165137614678899,
+      "grad_norm": 1.021941900253296,
+      "learning_rate": 9.805574477307118e-06,
+      "loss": 0.133,
+      "step": 690
+    },
+    {
+      "epoch": 3.188073394495413,
+      "grad_norm": 1.0381412506103516,
+      "learning_rate": 9.59261504624378e-06,
+      "loss": 0.1556,
+      "step": 695
+    },
+    {
+      "epoch": 3.2110091743119265,
+      "grad_norm": 0.9593328833580017,
+      "learning_rate": 9.380901098761319e-06,
+      "loss": 0.131,
+      "step": 700
+    },
+    {
+      "epoch": 3.2339449541284404,
+      "grad_norm": 0.8940817713737488,
+      "learning_rate": 9.170481398956927e-06,
+      "loss": 0.1014,
+      "step": 705
+    },
+    {
+      "epoch": 3.2568807339449544,
+      "grad_norm": 1.0616079568862915,
+      "learning_rate": 8.961404412823578e-06,
+      "loss": 0.1046,
+      "step": 710
+    },
+    {
+      "epoch": 3.279816513761468,
+      "grad_norm": 1.0666618347167969,
+      "learning_rate": 8.753718297086869e-06,
+      "loss": 0.1249,
+      "step": 715
+    },
+    {
+      "epoch": 3.302752293577982,
+      "grad_norm": 1.064950942993164,
+      "learning_rate": 8.547470888113046e-06,
+      "loss": 0.1089,
+      "step": 720
+    },
+    {
+      "epoch": 3.3256880733944953,
+      "grad_norm": 1.1599719524383545,
+      "learning_rate": 8.342709690890875e-06,
+      "loss": 0.1335,
+      "step": 725
+    },
+    {
+      "epoch": 3.3486238532110093,
+      "grad_norm": 1.2495425939559937,
+      "learning_rate": 8.139481868089797e-06,
+      "loss": 0.1377,
+      "step": 730
+    },
+    {
+      "epoch": 3.371559633027523,
+      "grad_norm": 1.13038170337677,
+      "learning_rate": 7.937834229196972e-06,
+      "loss": 0.1342,
+      "step": 735
+    },
+    {
+      "epoch": 3.3944954128440368,
+      "grad_norm": 1.1380192041397095,
+      "learning_rate": 7.7378132197356e-06,
+      "loss": 0.1387,
+      "step": 740
+    },
+    {
+      "epoch": 3.4174311926605503,
+      "grad_norm": 0.9990273714065552,
+      "learning_rate": 7.539464910567136e-06,
+      "loss": 0.1205,
+      "step": 745
+    },
+    {
+      "epoch": 3.4403669724770642,
+      "grad_norm": 1.0543484687805176,
+      "learning_rate": 7.342834987279823e-06,
+      "loss": 0.1234,
+      "step": 750
+    },
+    {
+      "epoch": 3.463302752293578,
+      "grad_norm": 0.9731931686401367,
+      "learning_rate": 7.147968739665866e-06,
+      "loss": 0.117,
+      "step": 755
+    },
+    {
+      "epoch": 3.4862385321100917,
+      "grad_norm": 0.9503082036972046,
+      "learning_rate": 6.954911051289854e-06,
+      "loss": 0.1192,
+      "step": 760
+    },
+    {
+      "epoch": 3.5091743119266052,
+      "grad_norm": 1.0039364099502563,
+      "learning_rate": 6.763706389150751e-06,
+      "loss": 0.1139,
+      "step": 765
+    },
+    {
+      "epoch": 3.532110091743119,
+      "grad_norm": 0.791185200214386,
+      "learning_rate": 6.574398793439745e-06,
+      "loss": 0.1064,
+      "step": 770
+    },
+    {
+      "epoch": 3.555045871559633,
+      "grad_norm": 1.1504929065704346,
+      "learning_rate": 6.387031867396516e-06,
+      "loss": 0.1214,
+      "step": 775
+    },
+    {
+      "epoch": 3.5779816513761467,
+      "grad_norm": 1.0110254287719727,
+      "learning_rate": 6.201648767266099e-06,
+      "loss": 0.1196,
+      "step": 780
+    },
+    {
+      "epoch": 3.6009174311926606,
+      "grad_norm": 0.9721340537071228,
+      "learning_rate": 6.018292192358662e-06,
+      "loss": 0.1028,
+      "step": 785
+    },
+    {
+      "epoch": 3.623853211009174,
+      "grad_norm": 0.8446089029312134,
+      "learning_rate": 5.837004375214604e-06,
+      "loss": 0.1063,
+      "step": 790
+    },
+    {
+      "epoch": 3.646788990825688,
+      "grad_norm": 1.0468168258666992,
+      "learning_rate": 5.65782707187711e-06,
+      "loss": 0.1178,
+      "step": 795
+    },
+    {
+      "epoch": 3.669724770642202,
+      "grad_norm": 1.2742372751235962,
+      "learning_rate": 5.4808015522744885e-06,
+      "loss": 0.1111,
+      "step": 800
+    },
+    {
+      "epoch": 3.6926605504587156,
+      "grad_norm": 1.0528957843780518,
+      "learning_rate": 5.3059685907144565e-06,
+      "loss": 0.0993,
+      "step": 805
+    },
+    {
+      "epoch": 3.7155963302752295,
+      "grad_norm": 1.0999575853347778,
+      "learning_rate": 5.133368456492599e-06,
+      "loss": 0.1141,
+      "step": 810
+    },
+    {
+      "epoch": 3.738532110091743,
+      "grad_norm": 0.9651297330856323,
+      "learning_rate": 4.963040904617131e-06,
+      "loss": 0.094,
+      "step": 815
+    },
+    {
+      "epoch": 3.761467889908257,
+      "grad_norm": 0.8356239199638367,
+      "learning_rate": 4.795025166652099e-06,
+      "loss": 0.1029,
+      "step": 820
+    },
+    {
+      "epoch": 3.7844036697247705,
+      "grad_norm": 1.0308773517608643,
+      "learning_rate": 4.629359941681226e-06,
+      "loss": 0.0944,
+      "step": 825
+    },
+    {
+      "epoch": 3.8073394495412844,
+      "grad_norm": 0.9902791976928711,
+      "learning_rate": 4.466083387394294e-06,
+      "loss": 0.1126,
+      "step": 830
+    },
+    {
+      "epoch": 3.830275229357798,
+      "grad_norm": 0.897495687007904,
+      "learning_rate": 4.305233111298303e-06,
+      "loss": 0.1192,
+      "step": 835
+    },
+    {
+      "epoch": 3.853211009174312,
+      "grad_norm": 0.8419690132141113,
+      "learning_rate": 4.146846162055325e-06,
+      "loss": 0.1001,
+      "step": 840
+    },
+    {
+      "epoch": 3.876146788990826,
+      "grad_norm": 0.9405183792114258,
+      "learning_rate": 3.990959020949096e-06,
+      "loss": 0.1058,
+      "step": 845
+    },
+    {
+      "epoch": 3.8990825688073394,
+      "grad_norm": 0.9448903799057007,
+      "learning_rate": 3.837607593482242e-06,
+      "loss": 0.0972,
+      "step": 850
+    },
+    {
+      "epoch": 3.9220183486238533,
+      "grad_norm": 0.9606824517250061,
+      "learning_rate": 3.6868272011061916e-06,
+      "loss": 0.0817,
+      "step": 855
+    },
+    {
+      "epoch": 3.944954128440367,
+      "grad_norm": 1.1081745624542236,
+      "learning_rate": 3.5386525730855875e-06,
+      "loss": 0.0973,
+      "step": 860
+    },
+    {
+      "epoch": 3.967889908256881,
+      "grad_norm": 0.8317539691925049,
+      "learning_rate": 3.39311783849906e-06,
+      "loss": 0.1002,
+      "step": 865
+    },
+    {
+      "epoch": 3.9908256880733948,
+      "grad_norm": 0.9665256142616272,
+      "learning_rate": 3.2502565183782956e-06,
+      "loss": 0.1053,
+      "step": 870
+    },
+    {
+      "epoch": 4.013761467889908,
+      "grad_norm": 0.7488636374473572,
+      "learning_rate": 3.110101517987129e-06,
+      "loss": 0.0779,
+      "step": 875
+    },
+    {
+      "epoch": 4.036697247706422,
+      "grad_norm": 0.8224223852157593,
+      "learning_rate": 2.9726851192424643e-06,
+      "loss": 0.0847,
+      "step": 880
+    },
+    {
+      "epoch": 4.059633027522936,
+      "grad_norm": 0.7991659641265869,
+      "learning_rate": 2.8380389732787865e-06,
+      "loss": 0.0827,
+      "step": 885
+    },
+    {
+      "epoch": 4.08256880733945,
+      "grad_norm": 0.665056586265564,
+      "learning_rate": 2.7061940931579527e-06,
+      "loss": 0.0798,
+      "step": 890
+    },
+    {
+      "epoch": 4.105504587155964,
+      "grad_norm": 0.6945267915725708,
+      "learning_rate": 2.577180846725939e-06,
+      "loss": 0.068,
+      "step": 895
+    },
+    {
+      "epoch": 4.128440366972477,
+      "grad_norm": 0.844860851764679,
+      "learning_rate": 2.4510289496182076e-06,
+      "loss": 0.0787,
+      "step": 900
+    },
+    {
+      "epoch": 4.151376146788991,
+      "grad_norm": 0.871397852897644,
+      "learning_rate": 2.3277674584153336e-06,
+      "loss": 0.0717,
+      "step": 905
+    },
+    {
+      "epoch": 4.174311926605505,
+      "grad_norm": 0.8091982007026672,
+      "learning_rate": 2.20742476395036e-06,
+      "loss": 0.0651,
+      "step": 910
+    },
+    {
+      "epoch": 4.197247706422019,
+      "grad_norm": 0.7804227471351624,
+      "learning_rate": 2.090028584769556e-06,
+      "loss": 0.0846,
+      "step": 915
+    },
+    {
+      "epoch": 4.220183486238533,
+      "grad_norm": 0.7627037763595581,
+      "learning_rate": 1.9756059607479964e-06,
+      "loss": 0.0703,
+      "step": 920
+    },
+    {
+      "epoch": 4.243119266055046,
+      "grad_norm": 0.7532461881637573,
+      "learning_rate": 1.8641832468614612e-06,
+      "loss": 0.088,
+      "step": 925
+    },
+    {
+      "epoch": 4.26605504587156,
+      "grad_norm": 0.9600714445114136,
+      "learning_rate": 1.7557861071160953e-06,
+      "loss": 0.0933,
+      "step": 930
+    },
+    {
+      "epoch": 4.2889908256880735,
+      "grad_norm": 0.7241982817649841,
+      "learning_rate": 1.6504395086372177e-06,
+      "loss": 0.0793,
+      "step": 935
+    },
+    {
+      "epoch": 4.3119266055045875,
+      "grad_norm": 0.786359429359436,
+      "learning_rate": 1.5481677159186309e-06,
+      "loss": 0.0841,
+      "step": 940
+    },
+    {
+      "epoch": 4.334862385321101,
+      "grad_norm": 0.8596308827400208,
+      "learning_rate": 1.4489942852337817e-06,
+      "loss": 0.0807,
+      "step": 945
+    },
+    {
+      "epoch": 4.3577981651376145,
+      "grad_norm": 0.7467135787010193,
+      "learning_rate": 1.35294205921005e-06,
+      "loss": 0.0663,
+      "step": 950
+    },
+    {
+      "epoch": 4.3807339449541285,
+      "grad_norm": 0.8038021326065063,
+      "learning_rate": 1.2600331615673898e-06,
+      "loss": 0.0805,
+      "step": 955
+    },
+    {
+      "epoch": 4.4036697247706424,
+      "grad_norm": 0.7694443464279175,
+      "learning_rate": 1.1702889920225835e-06,
+      "loss": 0.0882,
+      "step": 960
+    },
+    {
+      "epoch": 4.426605504587156,
+      "grad_norm": 0.759328305721283,
+      "learning_rate": 1.0837302213602312e-06,
+      "loss": 0.0698,
+      "step": 965
+    },
+    {
+      "epoch": 4.4495412844036695,
+      "grad_norm": 0.8927591443061829,
+      "learning_rate": 1.0003767866716646e-06,
+      "loss": 0.0798,
+      "step": 970
+    },
+    {
+      "epoch": 4.472477064220183,
+      "grad_norm": 0.854810357093811,
+      "learning_rate": 9.202478867628033e-07,
+      "loss": 0.0768,
+      "step": 975
+    },
+    {
+      "epoch": 4.495412844036697,
+      "grad_norm": 0.720846951007843,
+      "learning_rate": 8.433619777321222e-07,
+      "loss": 0.0752,
+      "step": 980
+    },
+    {
+      "epoch": 4.518348623853211,
+      "grad_norm": 0.7461158037185669,
+      "learning_rate": 7.697367687196488e-07,
+      "loss": 0.0808,
+      "step": 985
+    },
+    {
+      "epoch": 4.541284403669724,
+      "grad_norm": 0.7488417625427246,
+      "learning_rate": 6.99389217828002e-07,
+      "loss": 0.0663,
+      "step": 990
+    },
+    {
+      "epoch": 4.564220183486238,
+      "grad_norm": 0.782856285572052,
+      "learning_rate": 6.323355282164412e-07,
+      "loss": 0.0752,
+      "step": 995
+    },
+    {
+      "epoch": 4.587155963302752,
+      "grad_norm": 0.7895769476890564,
+      "learning_rate": 5.685911443687947e-07,
+      "loss": 0.0697,
+      "step": 1000
+    },
+    {
+      "epoch": 4.610091743119266,
+      "grad_norm": 0.8063625693321228,
+      "learning_rate": 5.081707485361159e-07,
+      "loss": 0.0725,
+      "step": 1005
+    },
+    {
+      "epoch": 4.63302752293578,
+      "grad_norm": 0.8060641288757324,
+      "learning_rate": 4.510882573549313e-07,
+      "loss": 0.0864,
+      "step": 1010
+    },
+    {
+      "epoch": 4.655963302752293,
+      "grad_norm": 0.7230294346809387,
+      "learning_rate": 3.9735681864181804e-07,
+      "loss": 0.0796,
+      "step": 1015
+    },
+    {
+      "epoch": 4.678899082568807,
+      "grad_norm": 0.8133239150047302,
+      "learning_rate": 3.4698880836506133e-07,
+      "loss": 0.0732,
+      "step": 1020
+    },
+    {
+      "epoch": 4.701834862385321,
+      "grad_norm": 0.6908076405525208,
+      "learning_rate": 2.999958277940978e-07,
+      "loss": 0.0702,
+      "step": 1025
+    },
+    {
+      "epoch": 4.724770642201835,
+      "grad_norm": 0.7993811964988708,
+      "learning_rate": 2.563887008273952e-07,
+      "loss": 0.0711,
+      "step": 1030
+    },
+    {
+      "epoch": 4.747706422018348,
+      "grad_norm": 0.6282556056976318,
+      "learning_rate": 2.161774714993825e-07,
+      "loss": 0.0746,
+      "step": 1035
+    },
+    {
+      "epoch": 4.770642201834862,
+      "grad_norm": 0.802282989025116,
+      "learning_rate": 1.7937140166700615e-07,
+      "loss": 0.0865,
+      "step": 1040
+    },
+    {
+      "epoch": 4.793577981651376,
+      "grad_norm": 0.6557889580726624,
+      "learning_rate": 1.4597896887644458e-07,
+      "loss": 0.0698,
+      "step": 1045
+    },
+    {
+      "epoch": 4.81651376146789,
+      "grad_norm": 0.684445858001709,
+      "learning_rate": 1.160078644104784e-07,
+      "loss": 0.0756,
+      "step": 1050
+    },
+    {
+      "epoch": 4.839449541284404,
+      "grad_norm": 0.7321560382843018,
+      "learning_rate": 8.946499151694598e-08,
+      "loss": 0.0639,
+      "step": 1055
+    },
+    {
+      "epoch": 4.862385321100917,
+      "grad_norm": 0.6626927256584167,
+      "learning_rate": 6.635646381872861e-08,
+      "loss": 0.0791,
+      "step": 1060
+    },
+    {
+      "epoch": 4.885321100917431,
+      "grad_norm": 0.7926541566848755,
+      "learning_rate": 4.6687603905595234e-08,
+      "loss": 0.0691,
+      "step": 1065
+    },
+    {
+      "epoch": 4.908256880733945,
+      "grad_norm": 0.6801010370254517,
+      "learning_rate": 3.046294210824973e-08,
+      "loss": 0.0814,
+      "step": 1070
+    },
+    {
+      "epoch": 4.931192660550459,
+      "grad_norm": 0.7773170471191406,
+      "learning_rate": 1.768621545485405e-08,
+      "loss": 0.076,
+      "step": 1075
+    },
+    {
+      "epoch": 4.954128440366972,
+      "grad_norm": 0.655495822429657,
+      "learning_rate": 8.360366810289954e-09,
+      "loss": 0.0782,
+      "step": 1080
+    },
+    {
+      "epoch": 4.977064220183486,
+      "grad_norm": 0.7868916392326355,
+      "learning_rate": 2.487544198316316e-09,
+      "loss": 0.0705,
+      "step": 1085
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.7025271058082581,
+      "learning_rate": 6.910030682150126e-11,
+      "loss": 0.0848,
+      "step": 1090
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1090,
+  "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": 1.5654684182351708e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

26_128_e5_3e-5/checkpoint-1090/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e8274742de25d720185c5eec61456a93760e442f86590eb1463c80adac56024
+size 7736

26_128_e5_3e-5/checkpoint-1090/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)

26_128_e5_3e-5/checkpoint-218/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

26_128_e5_3e-5/checkpoint-218/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

26_128_e5_3e-5/checkpoint-218/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9322cdd272447567e3afab7027bcd6e4b002dfd3829cbb3cc647ccd6a1dbd629
+size 791751704

26_128_e5_3e-5/checkpoint-218/latest

@@ -0,0 +1 @@
+global_step218

26_128_e5_3e-5/checkpoint-218/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:643bb2dcac13d7b97efffa514fc3f0e8b0317f2d48f87a5f3c998c22664e89fe
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bbdb57a80765e0172f3ec98a201c988337a5373fdd19450512c9c8064f91039
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f94704bcc785c1dea471d09a64e6b26b951f2bc0e27177199060c43c3d8f98e6
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59c2fc59c2064b6ea9276c906c429ff2b3072fe99405a04d7d9cac0e4fd7350c
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dae93ba14fee9474c10c2d0629b7e4a73998ddeae06da7b63ec78291f16ff03
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:547e68c1b1fde92c8d6e77855b42eeaddf4ef867a3e2219fc965ea2c2acf61f0
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:180541480171ade14ceed7ec41fd830ae5a59d1750c5c39c7780608904ed21f2
+size 15920

26_128_e5_3e-5/checkpoint-218/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0ba1db10ed222f485683cac8fa41897855687af194689b809a0075213f8b844
+size 15920

26_128_e5_3e-5/checkpoint-218/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3861418d7ca7e2a044f952a051f43a89df203b5aecad155ed9d0c9505f5ed928
+size 1064

26_128_e5_3e-5/checkpoint-218/special_tokens_map.json

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

26_128_e5_3e-5/checkpoint-218/tokenizer_config.json

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

26_128_e5_3e-5/checkpoint-218/trainer_state.json

@@ -0,0 +1,335 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 218,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.022935779816513763,
+      "grad_norm": 1.1547188758850098,
+      "learning_rate": 2.181818181818182e-06,
+      "loss": 1.3697,
+      "step": 5
+    },
+    {
+      "epoch": 0.045871559633027525,
+      "grad_norm": 0.7839906811714172,
+      "learning_rate": 4.90909090909091e-06,
+      "loss": 1.3117,
+      "step": 10
+    },
+    {
+      "epoch": 0.06880733944954129,
+      "grad_norm": 0.7896807193756104,
+      "learning_rate": 7.636363636363636e-06,
+      "loss": 1.3411,
+      "step": 15
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.4968399405479431,
+      "learning_rate": 1.0363636363636364e-05,
+      "loss": 1.3191,
+      "step": 20
+    },
+    {
+      "epoch": 0.11467889908256881,
+      "grad_norm": 0.48551270365715027,
+      "learning_rate": 1.309090909090909e-05,
+      "loss": 1.2715,
+      "step": 25
+    },
+    {
+      "epoch": 0.13761467889908258,
+      "grad_norm": 0.5304503440856934,
+      "learning_rate": 1.5818181818181818e-05,
+      "loss": 1.2378,
+      "step": 30
+    },
+    {
+      "epoch": 0.16055045871559634,
+      "grad_norm": 0.5629490613937378,
+      "learning_rate": 1.8545454545454545e-05,
+      "loss": 1.2796,
+      "step": 35
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.5019248723983765,
+      "learning_rate": 2.1272727272727273e-05,
+      "loss": 1.2179,
+      "step": 40
+    },
+    {
+      "epoch": 0.20642201834862386,
+      "grad_norm": 0.566190779209137,
+      "learning_rate": 2.4e-05,
+      "loss": 1.2287,
+      "step": 45
+    },
+    {
+      "epoch": 0.22935779816513763,
+      "grad_norm": 0.5070985555648804,
+      "learning_rate": 2.6727272727272728e-05,
+      "loss": 1.1711,
+      "step": 50
+    },
+    {
+      "epoch": 0.25229357798165136,
+      "grad_norm": 0.482239305973053,
+      "learning_rate": 2.9454545454545456e-05,
+      "loss": 1.127,
+      "step": 55
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.5259355306625366,
+      "learning_rate": 2.999889440782374e-05,
+      "loss": 1.1649,
+      "step": 60
+    },
+    {
+      "epoch": 0.2981651376146789,
+      "grad_norm": 0.48876306414604187,
+      "learning_rate": 2.9994403218928662e-05,
+      "loss": 1.1244,
+      "step": 65
+    },
+    {
+      "epoch": 0.3211009174311927,
+      "grad_norm": 0.5676949620246887,
+      "learning_rate": 2.9986458367463977e-05,
+      "loss": 1.0935,
+      "step": 70
+    },
+    {
+      "epoch": 0.3440366972477064,
+      "grad_norm": 0.5986928939819336,
+      "learning_rate": 2.997506168336822e-05,
+      "loss": 1.0729,
+      "step": 75
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.6229660511016846,
+      "learning_rate": 2.9960215791640898e-05,
+      "loss": 1.0235,
+      "step": 80
+    },
+    {
+      "epoch": 0.38990825688073394,
+      "grad_norm": 0.602012574672699,
+      "learning_rate": 2.9941924111737924e-05,
+      "loss": 0.9924,
+      "step": 85
+    },
+    {
+      "epoch": 0.41284403669724773,
+      "grad_norm": 0.6534562706947327,
+      "learning_rate": 2.9920190856783978e-05,
+      "loss": 1.0184,
+      "step": 90
+    },
+    {
+      "epoch": 0.43577981651376146,
+      "grad_norm": 0.6122053265571594,
+      "learning_rate": 2.98950210326021e-05,
+      "loss": 1.0002,
+      "step": 95
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.7329753041267395,
+      "learning_rate": 2.9866420436560697e-05,
+      "loss": 0.9922,
+      "step": 100
+    },
+    {
+      "epoch": 0.481651376146789,
+      "grad_norm": 0.7041805386543274,
+      "learning_rate": 2.9834395656238264e-05,
+      "loss": 0.9824,
+      "step": 105
+    },
+    {
+      "epoch": 0.5045871559633027,
+      "grad_norm": 0.6798579692840576,
+      "learning_rate": 2.979895406790603e-05,
+      "loss": 0.9195,
+      "step": 110
+    },
+    {
+      "epoch": 0.5275229357798165,
+      "grad_norm": 0.6428303122520447,
+      "learning_rate": 2.9760103834828996e-05,
+      "loss": 0.9656,
+      "step": 115
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.7067289352416992,
+      "learning_rate": 2.971785390538569e-05,
+      "loss": 0.8735,
+      "step": 120
+    },
+    {
+      "epoch": 0.573394495412844,
+      "grad_norm": 0.9809579253196716,
+      "learning_rate": 2.9672214011007087e-05,
+      "loss": 0.8944,
+      "step": 125
+    },
+    {
+      "epoch": 0.5963302752293578,
+      "grad_norm": 0.7807005643844604,
+      "learning_rate": 2.9623194663935153e-05,
+      "loss": 0.8463,
+      "step": 130
+    },
+    {
+      "epoch": 0.6192660550458715,
+      "grad_norm": 0.8913285136222839,
+      "learning_rate": 2.9570807154801587e-05,
+      "loss": 0.8572,
+      "step": 135
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.9661921262741089,
+      "learning_rate": 2.951506355002722e-05,
+      "loss": 0.8317,
+      "step": 140
+    },
+    {
+      "epoch": 0.6651376146788991,
+      "grad_norm": 0.785257875919342,
+      "learning_rate": 2.9455976689042774e-05,
+      "loss": 0.7877,
+      "step": 145
+    },
+    {
+      "epoch": 0.6880733944954128,
+      "grad_norm": 0.8387299180030823,
+      "learning_rate": 2.9393560181331543e-05,
+      "loss": 0.7675,
+      "step": 150
+    },
+    {
+      "epoch": 0.7110091743119266,
+      "grad_norm": 0.83286452293396,
+      "learning_rate": 2.9327828403294725e-05,
+      "loss": 0.767,
+      "step": 155
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.97318035364151,
+      "learning_rate": 2.92587964949401e-05,
+      "loss": 0.7821,
+      "step": 160
+    },
+    {
+      "epoch": 0.7568807339449541,
+      "grad_norm": 1.0256941318511963,
+      "learning_rate": 2.918648035639484e-05,
+      "loss": 0.7215,
+      "step": 165
+    },
+    {
+      "epoch": 0.7798165137614679,
+      "grad_norm": 0.8714763522148132,
+      "learning_rate": 2.9110896644243226e-05,
+      "loss": 0.7473,
+      "step": 170
+    },
+    {
+      "epoch": 0.8027522935779816,
+      "grad_norm": 0.914694607257843,
+      "learning_rate": 2.9032062767690116e-05,
+      "loss": 0.7344,
+      "step": 175
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 1.7585707902908325,
+      "learning_rate": 2.8949996884551118e-05,
+      "loss": 0.676,
+      "step": 180
+    },
+    {
+      "epoch": 0.8486238532110092,
+      "grad_norm": 0.9376805424690247,
+      "learning_rate": 2.886471789707026e-05,
+      "loss": 0.7359,
+      "step": 185
+    },
+    {
+      "epoch": 0.8715596330275229,
+      "grad_norm": 0.948411226272583,
+      "learning_rate": 2.8776245447566274e-05,
+      "loss": 0.6896,
+      "step": 190
+    },
+    {
+      "epoch": 0.8944954128440367,
+      "grad_norm": 1.0342053174972534,
+      "learning_rate": 2.868459991390833e-05,
+      "loss": 0.688,
+      "step": 195
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 1.0501846075057983,
+      "learning_rate": 2.8589802404822457e-05,
+      "loss": 0.6407,
+      "step": 200
+    },
+    {
+      "epoch": 0.9403669724770642,
+      "grad_norm": 0.932583212852478,
+      "learning_rate": 2.8491874755029502e-05,
+      "loss": 0.6745,
+      "step": 205
+    },
+    {
+      "epoch": 0.963302752293578,
+      "grad_norm": 1.0710359811782837,
+      "learning_rate": 2.8390839520216004e-05,
+      "loss": 0.6641,
+      "step": 210
+    },
+    {
+      "epoch": 0.9862385321100917,
+      "grad_norm": 1.0051730871200562,
+      "learning_rate": 2.8286719971838885e-05,
+      "loss": 0.6803,
+      "step": 215
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1090,
+  "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": 3.141616696242995e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

26_128_e5_3e-5/checkpoint-218/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e8274742de25d720185c5eec61456a93760e442f86590eb1463c80adac56024
+size 7736

26_128_e5_3e-5/checkpoint-218/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)

26_128_e5_3e-5/checkpoint-436/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

26_128_e5_3e-5/checkpoint-436/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": [
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

26_128_e5_3e-5/checkpoint-436/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:498c69abd087e68ca71ba61218776a3a49bdc3f7953c5bcd05b9aeb6f5700cb6
+size 791751704

26_128_e5_3e-5/checkpoint-436/latest

@@ -0,0 +1 @@
+global_step436

26_128_e5_3e-5/checkpoint-436/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6e66e828e165b829b74b21907a3b9136567559cf25669f504044ed14be232a4
+size 15920

26_128_e5_3e-5/checkpoint-436/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:727f9cb077f0e9cc8cf52206948f249b50f43957c6ac58b6451ea9ed626d1f3c
+size 15920

26_128_e5_3e-5/checkpoint-436/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad882ea4030ff6c3847a46a4c2bff89d88034f0a35e5f56292b394ccc13edb86
+size 15920