Upload folder using huggingface_hub

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/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 0.12.0

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "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_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "k_proj",
+    "v_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7376a510876af63e46d950223de15a48382fd21155c4d932913ba22de24948ff
+size 791751704

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/latest

@@ -0,0 +1 @@
+global_step513

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34ce998dc90e5b58ee264dc92a7011c1691aba3da1f900b1766b5910e4bd14fc
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acb67a9355c87cb188dd7a915e33581e98489567c2d23b643c994f7e53ea555f
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a5d518e80648b2ae9ad53b74f6b9554edb2abac53f69f8b9ec1a1a41b54ea33
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c31602cb60cec4148427538c27f3c82d1cf696c2d380dcf625e27f42b2fb60a1
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1c7eb8bfa2e654e2a0f2b1a78e763e5b2bd5ce627cc8dd6711e9dacd7524da0
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3c625f07536592873d52db06f011f77a81c4118aa25707be94b8c3bea7dc7f9
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25c8cf5aca85e630bd92ef136e47cde157a64420d4cefe4a0150a97e7a39c891
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:482a2afa3704fd56a9ea56238fb3cddcd927c5873325b6dfbcac269b51a60f85
+size 15920

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08bb599f8f80d068049d45be04c8657de2a8ecc9cfa9fadcb4c0e6bf9802d841
+size 1064

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/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
+  }
+}

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/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
+}

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/trainer_state.json

@@ -0,0 +1,748 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 514,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01949317738791423,
+      "grad_norm": 0.9854941964149475,
+      "learning_rate": 1.875e-06,
+      "loss": 1.3535,
+      "step": 5
+    },
+    {
+      "epoch": 0.03898635477582846,
+      "grad_norm": 0.8216362595558167,
+      "learning_rate": 4.21875e-06,
+      "loss": 1.3855,
+      "step": 10
+    },
+    {
+      "epoch": 0.05847953216374269,
+      "grad_norm": 0.47613006830215454,
+      "learning_rate": 6.5625e-06,
+      "loss": 1.427,
+      "step": 15
+    },
+    {
+      "epoch": 0.07797270955165692,
+      "grad_norm": 0.43144744634628296,
+      "learning_rate": 8.90625e-06,
+      "loss": 1.3401,
+      "step": 20
+    },
+    {
+      "epoch": 0.09746588693957114,
+      "grad_norm": 0.4169255793094635,
+      "learning_rate": 1.125e-05,
+      "loss": 1.3386,
+      "step": 25
+    },
+    {
+      "epoch": 0.11695906432748537,
+      "grad_norm": 0.43670251965522766,
+      "learning_rate": 1.359375e-05,
+      "loss": 1.3226,
+      "step": 30
+    },
+    {
+      "epoch": 0.1364522417153996,
+      "grad_norm": 0.46294134855270386,
+      "learning_rate": 1.59375e-05,
+      "loss": 1.269,
+      "step": 35
+    },
+    {
+      "epoch": 0.15594541910331383,
+      "grad_norm": 0.4115634858608246,
+      "learning_rate": 1.828125e-05,
+      "loss": 1.3314,
+      "step": 40
+    },
+    {
+      "epoch": 0.17543859649122806,
+      "grad_norm": 0.4386284053325653,
+      "learning_rate": 2.0625e-05,
+      "loss": 1.3226,
+      "step": 45
+    },
+    {
+      "epoch": 0.1949317738791423,
+      "grad_norm": 0.43195125460624695,
+      "learning_rate": 2.296875e-05,
+      "loss": 1.2715,
+      "step": 50
+    },
+    {
+      "epoch": 0.21442495126705652,
+      "grad_norm": 0.3740931451320648,
+      "learning_rate": 2.5312500000000002e-05,
+      "loss": 1.2907,
+      "step": 55
+    },
+    {
+      "epoch": 0.23391812865497075,
+      "grad_norm": 0.37740710377693176,
+      "learning_rate": 2.765625e-05,
+      "loss": 1.2552,
+      "step": 60
+    },
+    {
+      "epoch": 0.253411306042885,
+      "grad_norm": 0.36451616883277893,
+      "learning_rate": 3e-05,
+      "loss": 1.2549,
+      "step": 65
+    },
+    {
+      "epoch": 0.2729044834307992,
+      "grad_norm": 0.4989257752895355,
+      "learning_rate": 2.9998748508718575e-05,
+      "loss": 1.2123,
+      "step": 70
+    },
+    {
+      "epoch": 0.29239766081871343,
+      "grad_norm": 0.4312180280685425,
+      "learning_rate": 2.9994994243705013e-05,
+      "loss": 1.2314,
+      "step": 75
+    },
+    {
+      "epoch": 0.31189083820662766,
+      "grad_norm": 0.397249698638916,
+      "learning_rate": 2.9988737831416642e-05,
+      "loss": 1.2402,
+      "step": 80
+    },
+    {
+      "epoch": 0.3313840155945419,
+      "grad_norm": 0.47213783860206604,
+      "learning_rate": 2.9979980315832853e-05,
+      "loss": 1.1228,
+      "step": 85
+    },
+    {
+      "epoch": 0.3508771929824561,
+      "grad_norm": 0.46933478116989136,
+      "learning_rate": 2.9968723158280906e-05,
+      "loss": 1.1763,
+      "step": 90
+    },
+    {
+      "epoch": 0.37037037037037035,
+      "grad_norm": 0.5299496650695801,
+      "learning_rate": 2.995496823719206e-05,
+      "loss": 1.1305,
+      "step": 95
+    },
+    {
+      "epoch": 0.3898635477582846,
+      "grad_norm": 0.8233343362808228,
+      "learning_rate": 2.9938717847788167e-05,
+      "loss": 1.1215,
+      "step": 100
+    },
+    {
+      "epoch": 0.4093567251461988,
+      "grad_norm": 0.6264380812644958,
+      "learning_rate": 2.9919974701698638e-05,
+      "loss": 1.1106,
+      "step": 105
+    },
+    {
+      "epoch": 0.42884990253411304,
+      "grad_norm": 0.5522003173828125,
+      "learning_rate": 2.989874192650801e-05,
+      "loss": 1.0699,
+      "step": 110
+    },
+    {
+      "epoch": 0.44834307992202727,
+      "grad_norm": 0.5564708113670349,
+      "learning_rate": 2.9875023065234003e-05,
+      "loss": 1.1112,
+      "step": 115
+    },
+    {
+      "epoch": 0.4678362573099415,
+      "grad_norm": 0.5427656173706055,
+      "learning_rate": 2.984882207573638e-05,
+      "loss": 1.112,
+      "step": 120
+    },
+    {
+      "epoch": 0.4873294346978557,
+      "grad_norm": 1.1721397638320923,
+      "learning_rate": 2.982014333005645e-05,
+      "loss": 1.0307,
+      "step": 125
+    },
+    {
+      "epoch": 0.50682261208577,
+      "grad_norm": 3.603581666946411,
+      "learning_rate": 2.9788991613687575e-05,
+      "loss": 1.0366,
+      "step": 130
+    },
+    {
+      "epoch": 0.5263157894736842,
+      "grad_norm": 0.7015699744224548,
+      "learning_rate": 2.9755372124776616e-05,
+      "loss": 1.0402,
+      "step": 135
+    },
+    {
+      "epoch": 0.5458089668615984,
+      "grad_norm": 0.5572109818458557,
+      "learning_rate": 2.971929047325654e-05,
+      "loss": 1.0234,
+      "step": 140
+    },
+    {
+      "epoch": 0.5653021442495126,
+      "grad_norm": 0.7717771530151367,
+      "learning_rate": 2.968075267991032e-05,
+      "loss": 1.0566,
+      "step": 145
+    },
+    {
+      "epoch": 0.5847953216374269,
+      "grad_norm": 0.6046805381774902,
+      "learning_rate": 2.963976517536627e-05,
+      "loss": 0.9953,
+      "step": 150
+    },
+    {
+      "epoch": 0.6042884990253411,
+      "grad_norm": 0.6307802200317383,
+      "learning_rate": 2.9596334799025003e-05,
+      "loss": 1.0418,
+      "step": 155
+    },
+    {
+      "epoch": 0.6237816764132553,
+      "grad_norm": 0.6888144612312317,
+      "learning_rate": 2.9550468797918162e-05,
+      "loss": 0.9906,
+      "step": 160
+    },
+    {
+      "epoch": 0.6432748538011696,
+      "grad_norm": 0.7008697986602783,
+      "learning_rate": 2.950217482549915e-05,
+      "loss": 0.973,
+      "step": 165
+    },
+    {
+      "epoch": 0.6627680311890838,
+      "grad_norm": 0.6218302249908447,
+      "learning_rate": 2.9451460940366025e-05,
+      "loss": 0.9422,
+      "step": 170
+    },
+    {
+      "epoch": 0.682261208576998,
+      "grad_norm": 0.6840229630470276,
+      "learning_rate": 2.9398335604916797e-05,
+      "loss": 0.9563,
+      "step": 175
+    },
+    {
+      "epoch": 0.7017543859649122,
+      "grad_norm": 0.7676129341125488,
+      "learning_rate": 2.9342807683937352e-05,
+      "loss": 0.9923,
+      "step": 180
+    },
+    {
+      "epoch": 0.7212475633528265,
+      "grad_norm": 0.7137411236763,
+      "learning_rate": 2.928488644312222e-05,
+      "loss": 0.9973,
+      "step": 185
+    },
+    {
+      "epoch": 0.7407407407407407,
+      "grad_norm": 0.7639793157577515,
+      "learning_rate": 2.9224581547528453e-05,
+      "loss": 0.9568,
+      "step": 190
+    },
+    {
+      "epoch": 0.7602339181286549,
+      "grad_norm": 0.7395185232162476,
+      "learning_rate": 2.916190305996286e-05,
+      "loss": 0.9299,
+      "step": 195
+    },
+    {
+      "epoch": 0.7797270955165692,
+      "grad_norm": 0.7606521844863892,
+      "learning_rate": 2.909686143930287e-05,
+      "loss": 0.9247,
+      "step": 200
+    },
+    {
+      "epoch": 0.7992202729044834,
+      "grad_norm": 0.779329776763916,
+      "learning_rate": 2.902946753875131e-05,
+      "loss": 0.906,
+      "step": 205
+    },
+    {
+      "epoch": 0.8187134502923976,
+      "grad_norm": 0.7310610413551331,
+      "learning_rate": 2.895973260402537e-05,
+      "loss": 0.8739,
+      "step": 210
+    },
+    {
+      "epoch": 0.8382066276803118,
+      "grad_norm": 0.8392149806022644,
+      "learning_rate": 2.8887668271480098e-05,
+      "loss": 0.8959,
+      "step": 215
+    },
+    {
+      "epoch": 0.8576998050682261,
+      "grad_norm": 1.0112305879592896,
+      "learning_rate": 2.8813286566166674e-05,
+      "loss": 0.9094,
+      "step": 220
+    },
+    {
+      "epoch": 0.8771929824561403,
+      "grad_norm": 0.8116384744644165,
+      "learning_rate": 2.873659989982586e-05,
+      "loss": 0.8674,
+      "step": 225
+    },
+    {
+      "epoch": 0.8966861598440545,
+      "grad_norm": 0.8890613913536072,
+      "learning_rate": 2.8657621068816903e-05,
+      "loss": 0.8724,
+      "step": 230
+    },
+    {
+      "epoch": 0.9161793372319688,
+      "grad_norm": 0.901253342628479,
+      "learning_rate": 2.857636325198225e-05,
+      "loss": 0.8044,
+      "step": 235
+    },
+    {
+      "epoch": 0.935672514619883,
+      "grad_norm": 0.8452696800231934,
+      "learning_rate": 2.849284000844849e-05,
+      "loss": 0.8755,
+      "step": 240
+    },
+    {
+      "epoch": 0.9551656920077972,
+      "grad_norm": 1.0774387121200562,
+      "learning_rate": 2.8407065275363756e-05,
+      "loss": 0.8576,
+      "step": 245
+    },
+    {
+      "epoch": 0.9746588693957114,
+      "grad_norm": 0.8810860514640808,
+      "learning_rate": 2.8319053365572135e-05,
+      "loss": 0.8204,
+      "step": 250
+    },
+    {
+      "epoch": 0.9941520467836257,
+      "grad_norm": 0.8354170322418213,
+      "learning_rate": 2.8228818965225325e-05,
+      "loss": 0.8506,
+      "step": 255
+    },
+    {
+      "epoch": 1.0116959064327484,
+      "grad_norm": 0.9805569648742676,
+      "learning_rate": 2.8136377131332043e-05,
+      "loss": 0.7696,
+      "step": 260
+    },
+    {
+      "epoch": 1.0311890838206628,
+      "grad_norm": 0.8466799855232239,
+      "learning_rate": 2.8041743289245503e-05,
+      "loss": 0.7292,
+      "step": 265
+    },
+    {
+      "epoch": 1.050682261208577,
+      "grad_norm": 0.8261882662773132,
+      "learning_rate": 2.7944933230089484e-05,
+      "loss": 0.7287,
+      "step": 270
+    },
+    {
+      "epoch": 1.0701754385964912,
+      "grad_norm": 0.9157257080078125,
+      "learning_rate": 2.784596310812331e-05,
+      "loss": 0.7328,
+      "step": 275
+    },
+    {
+      "epoch": 1.0896686159844053,
+      "grad_norm": 0.8392409682273865,
+      "learning_rate": 2.774484943804629e-05,
+      "loss": 0.7363,
+      "step": 280
+    },
+    {
+      "epoch": 1.1091617933723197,
+      "grad_norm": 0.8547697067260742,
+      "learning_rate": 2.764160909224196e-05,
+      "loss": 0.7384,
+      "step": 285
+    },
+    {
+      "epoch": 1.128654970760234,
+      "grad_norm": 0.8612291812896729,
+      "learning_rate": 2.7536259297962674e-05,
+      "loss": 0.7273,
+      "step": 290
+    },
+    {
+      "epoch": 1.1481481481481481,
+      "grad_norm": 0.8668325543403625,
+      "learning_rate": 2.7428817634454973e-05,
+      "loss": 0.7188,
+      "step": 295
+    },
+    {
+      "epoch": 1.1676413255360623,
+      "grad_norm": 0.931692898273468,
+      "learning_rate": 2.7319302030026207e-05,
+      "loss": 0.7064,
+      "step": 300
+    },
+    {
+      "epoch": 1.1871345029239766,
+      "grad_norm": 1.0540717840194702,
+      "learning_rate": 2.7207730759052925e-05,
+      "loss": 0.6855,
+      "step": 305
+    },
+    {
+      "epoch": 1.206627680311891,
+      "grad_norm": 1.041435718536377,
+      "learning_rate": 2.7094122438931513e-05,
+      "loss": 0.6789,
+      "step": 310
+    },
+    {
+      "epoch": 1.226120857699805,
+      "grad_norm": 0.9543320536613464,
+      "learning_rate": 2.697849602697159e-05,
+      "loss": 0.6979,
+      "step": 315
+    },
+    {
+      "epoch": 1.2456140350877192,
+      "grad_norm": 1.0173633098602295,
+      "learning_rate": 2.6860870817232682e-05,
+      "loss": 0.6924,
+      "step": 320
+    },
+    {
+      "epoch": 1.2651072124756335,
+      "grad_norm": 0.8906193375587463,
+      "learning_rate": 2.6741266437304718e-05,
+      "loss": 0.6976,
+      "step": 325
+    },
+    {
+      "epoch": 1.2846003898635479,
+      "grad_norm": 1.3355520963668823,
+      "learning_rate": 2.661970284503286e-05,
+      "loss": 0.646,
+      "step": 330
+    },
+    {
+      "epoch": 1.304093567251462,
+      "grad_norm": 1.012062668800354,
+      "learning_rate": 2.6496200325187223e-05,
+      "loss": 0.6911,
+      "step": 335
+    },
+    {
+      "epoch": 1.323586744639376,
+      "grad_norm": 0.9941599369049072,
+      "learning_rate": 2.6370779486078047e-05,
+      "loss": 0.6703,
+      "step": 340
+    },
+    {
+      "epoch": 1.3430799220272904,
+      "grad_norm": 1.0016067028045654,
+      "learning_rate": 2.6243461256116892e-05,
+      "loss": 0.6903,
+      "step": 345
+    },
+    {
+      "epoch": 1.3625730994152048,
+      "grad_norm": 0.8689593076705933,
+      "learning_rate": 2.611426688032439e-05,
+      "loss": 0.6504,
+      "step": 350
+    },
+    {
+      "epoch": 1.3820662768031189,
+      "grad_norm": 0.8714726567268372,
+      "learning_rate": 2.598321791678519e-05,
+      "loss": 0.6562,
+      "step": 355
+    },
+    {
+      "epoch": 1.401559454191033,
+      "grad_norm": 0.9304203987121582,
+      "learning_rate": 2.5850336233050677e-05,
+      "loss": 0.6671,
+      "step": 360
+    },
+    {
+      "epoch": 1.4210526315789473,
+      "grad_norm": 1.0773100852966309,
+      "learning_rate": 2.5715644002489998e-05,
+      "loss": 0.6376,
+      "step": 365
+    },
+    {
+      "epoch": 1.4405458089668617,
+      "grad_norm": 0.96257483959198,
+      "learning_rate": 2.557916370059012e-05,
+      "loss": 0.6263,
+      "step": 370
+    },
+    {
+      "epoch": 1.4600389863547758,
+      "grad_norm": 0.9565174579620361,
+      "learning_rate": 2.544091810120543e-05,
+      "loss": 0.6413,
+      "step": 375
+    },
+    {
+      "epoch": 1.47953216374269,
+      "grad_norm": 1.0495936870574951,
+      "learning_rate": 2.530093027275757e-05,
+      "loss": 0.6373,
+      "step": 380
+    },
+    {
+      "epoch": 1.4990253411306043,
+      "grad_norm": 1.0575157403945923,
+      "learning_rate": 2.5159223574386117e-05,
+      "loss": 0.6373,
+      "step": 385
+    },
+    {
+      "epoch": 1.5185185185185186,
+      "grad_norm": 1.2348133325576782,
+      "learning_rate": 2.501582165205074e-05,
+      "loss": 0.6296,
+      "step": 390
+    },
+    {
+      "epoch": 1.5380116959064327,
+      "grad_norm": 0.9543783664703369,
+      "learning_rate": 2.4870748434585514e-05,
+      "loss": 0.6238,
+      "step": 395
+    },
+    {
+      "epoch": 1.5575048732943468,
+      "grad_norm": 1.031792163848877,
+      "learning_rate": 2.4724028129706012e-05,
+      "loss": 0.6156,
+      "step": 400
+    },
+    {
+      "epoch": 1.5769980506822612,
+      "grad_norm": 1.0421391725540161,
+      "learning_rate": 2.4575685219969884e-05,
+      "loss": 0.6188,
+      "step": 405
+    },
+    {
+      "epoch": 1.5964912280701755,
+      "grad_norm": 1.0755267143249512,
+      "learning_rate": 2.442574445869156e-05,
+      "loss": 0.5753,
+      "step": 410
+    },
+    {
+      "epoch": 1.6159844054580896,
+      "grad_norm": 0.9432569146156311,
+      "learning_rate": 2.4274230865811763e-05,
+      "loss": 0.5878,
+      "step": 415
+    },
+    {
+      "epoch": 1.6354775828460038,
+      "grad_norm": 1.234410047531128,
+      "learning_rate": 2.4121169723722566e-05,
+      "loss": 0.6097,
+      "step": 420
+    },
+    {
+      "epoch": 1.654970760233918,
+      "grad_norm": 1.4969490766525269,
+      "learning_rate": 2.396658657304861e-05,
+      "loss": 0.5825,
+      "step": 425
+    },
+    {
+      "epoch": 1.6744639376218324,
+      "grad_norm": 1.0849074125289917,
+      "learning_rate": 2.381050720838528e-05,
+      "loss": 0.5705,
+      "step": 430
+    },
+    {
+      "epoch": 1.6939571150097466,
+      "grad_norm": 1.682335376739502,
+      "learning_rate": 2.3652957673994448e-05,
+      "loss": 0.5889,
+      "step": 435
+    },
+    {
+      "epoch": 1.7134502923976607,
+      "grad_norm": 1.1412495374679565,
+      "learning_rate": 2.3493964259458603e-05,
+      "loss": 0.5789,
+      "step": 440
+    },
+    {
+      "epoch": 1.732943469785575,
+      "grad_norm": 1.067293405532837,
+      "learning_rate": 2.3333553495294033e-05,
+      "loss": 0.5971,
+      "step": 445
+    },
+    {
+      "epoch": 1.7524366471734893,
+      "grad_norm": 1.291559100151062,
+      "learning_rate": 2.317175214852377e-05,
+      "loss": 0.577,
+      "step": 450
+    },
+    {
+      "epoch": 1.7719298245614035,
+      "grad_norm": 1.2236227989196777,
+      "learning_rate": 2.3008587218211127e-05,
+      "loss": 0.578,
+      "step": 455
+    },
+    {
+      "epoch": 1.7914230019493176,
+      "grad_norm": 1.7148844003677368,
+      "learning_rate": 2.284408593095446e-05,
+      "loss": 0.562,
+      "step": 460
+    },
+    {
+      "epoch": 1.810916179337232,
+      "grad_norm": 1.1242623329162598,
+      "learning_rate": 2.2678275736344014e-05,
+      "loss": 0.5407,
+      "step": 465
+    },
+    {
+      "epoch": 1.8304093567251463,
+      "grad_norm": 1.1884270906448364,
+      "learning_rate": 2.251118430238151e-05,
+      "loss": 0.5363,
+      "step": 470
+    },
+    {
+      "epoch": 1.8499025341130604,
+      "grad_norm": 0.9591031074523926,
+      "learning_rate": 2.2342839510863323e-05,
+      "loss": 0.5671,
+      "step": 475
+    },
+    {
+      "epoch": 1.8693957115009745,
+      "grad_norm": 1.0892689228057861,
+      "learning_rate": 2.2173269452727965e-05,
+      "loss": 0.5199,
+      "step": 480
+    },
+    {
+      "epoch": 1.8888888888888888,
+      "grad_norm": 0.9958884119987488,
+      "learning_rate": 2.200250242336868e-05,
+      "loss": 0.5413,
+      "step": 485
+    },
+    {
+      "epoch": 1.9083820662768032,
+      "grad_norm": 1.1685152053833008,
+      "learning_rate": 2.183056691791193e-05,
+      "loss": 0.5257,
+      "step": 490
+    },
+    {
+      "epoch": 1.9278752436647173,
+      "grad_norm": 1.0781997442245483,
+      "learning_rate": 2.1657491626462514e-05,
+      "loss": 0.5417,
+      "step": 495
+    },
+    {
+      "epoch": 1.9473684210526314,
+      "grad_norm": 0.943932831287384,
+      "learning_rate": 2.1483305429316208e-05,
+      "loss": 0.4997,
+      "step": 500
+    },
+    {
+      "epoch": 1.9668615984405458,
+      "grad_norm": 1.057703971862793,
+      "learning_rate": 2.1308037392140613e-05,
+      "loss": 0.4959,
+      "step": 505
+    },
+    {
+      "epoch": 1.98635477582846,
+      "grad_norm": 1.0127170085906982,
+      "learning_rate": 2.113171676112513e-05,
+      "loss": 0.4931,
+      "step": 510
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1280,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1804183649299988e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a098ffc6270ef4911bfdb2d22f94de302c6b671bdf91a3bc5cd3ec9d81e76307
+size 7800

hotpotqa/not_grounded/0_128_e5_3e-5/checkpoint-514/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)

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/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 0.12.0

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "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_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": [
+    "down_proj",
+    "q_proj",
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b6119bcb8aeee90a738af9497970f903813c1649f405c716a2788fe43b6f469
+size 791751704

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/latest

@@ -0,0 +1 @@
+global_step642

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f88045bd6d82193f8e4514187bbb2798651676524ea226b9651428019611790
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ef4704c482539951921fcb331a8ca065431e2569b4d905a7b496266aa840b65
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8986ab3b6dade0671fcd0d6c18c7b4e8735570167545f0506826d532d90222db
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d1c594903c8cfcfe8e0993c14f8381d628a4e157d0043840d7567b815cbe1bd
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dd2fd7d673f3fa4d7b08afbda1e16253f604cc318ef5f736132879b51821f04
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7778d14b4c559506c1bc109ea24464ceea5268caaa5a85580834dbb8a08e0e6f
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b639297c84b275df5df2d93dc9ba444819a9fc75463ee50cbb8232ca0717bdc
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:035640cb191020272db90eb4c468805c56d39f3667d828d9a6ea82ad7f3a50c3
+size 15920

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7d2474414aabe8a8cf885ef31a66f3e6221d30ffc8db5478542d6dfbb253e31
+size 1064

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/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
+  }
+}

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/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
+}

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/trainer_state.json

@@ -0,0 +1,930 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 642,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02336448598130841,
+      "grad_norm": 0.9846668243408203,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 1.3054,
+      "step": 5
+    },
+    {
+      "epoch": 0.04672897196261682,
+      "grad_norm": 0.6714785099029541,
+      "learning_rate": 4.9999999999999996e-06,
+      "loss": 1.3098,
+      "step": 10
+    },
+    {
+      "epoch": 0.07009345794392523,
+      "grad_norm": 0.5717726945877075,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 1.2853,
+      "step": 15
+    },
+    {
+      "epoch": 0.09345794392523364,
+      "grad_norm": 0.4370105266571045,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 1.2847,
+      "step": 20
+    },
+    {
+      "epoch": 0.11682242990654206,
+      "grad_norm": 0.47119808197021484,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.2456,
+      "step": 25
+    },
+    {
+      "epoch": 0.14018691588785046,
+      "grad_norm": 0.3612341582775116,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 1.2906,
+      "step": 30
+    },
+    {
+      "epoch": 0.16355140186915887,
+      "grad_norm": 0.4175552427768707,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 1.2321,
+      "step": 35
+    },
+    {
+      "epoch": 0.18691588785046728,
+      "grad_norm": 0.3725614547729492,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 1.2422,
+      "step": 40
+    },
+    {
+      "epoch": 0.2102803738317757,
+      "grad_norm": 0.3745448887348175,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 1.166,
+      "step": 45
+    },
+    {
+      "epoch": 0.2336448598130841,
+      "grad_norm": 0.3536573648452759,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 1.1849,
+      "step": 50
+    },
+    {
+      "epoch": 0.2570093457943925,
+      "grad_norm": 0.376469224691391,
+      "learning_rate": 3e-05,
+      "loss": 1.1594,
+      "step": 55
+    },
+    {
+      "epoch": 0.2803738317757009,
+      "grad_norm": 0.40825456380844116,
+      "learning_rate": 2.9998207311011924e-05,
+      "loss": 1.1643,
+      "step": 60
+    },
+    {
+      "epoch": 0.3037383177570093,
+      "grad_norm": 0.4157944917678833,
+      "learning_rate": 2.9992829672545544e-05,
+      "loss": 1.1724,
+      "step": 65
+    },
+    {
+      "epoch": 0.32710280373831774,
+      "grad_norm": 0.40997833013534546,
+      "learning_rate": 2.9983868369991955e-05,
+      "loss": 1.1053,
+      "step": 70
+    },
+    {
+      "epoch": 0.35046728971962615,
+      "grad_norm": 0.4295913279056549,
+      "learning_rate": 2.997132554532828e-05,
+      "loss": 1.1017,
+      "step": 75
+    },
+    {
+      "epoch": 0.37383177570093457,
+      "grad_norm": 0.4532783329486847,
+      "learning_rate": 2.9955204196605673e-05,
+      "loss": 1.0855,
+      "step": 80
+    },
+    {
+      "epoch": 0.397196261682243,
+      "grad_norm": 0.504647433757782,
+      "learning_rate": 2.9935508177232713e-05,
+      "loss": 1.0616,
+      "step": 85
+    },
+    {
+      "epoch": 0.4205607476635514,
+      "grad_norm": 0.457560658454895,
+      "learning_rate": 2.9912242195054337e-05,
+      "loss": 1.0598,
+      "step": 90
+    },
+    {
+      "epoch": 0.4439252336448598,
+      "grad_norm": 0.46719983220100403,
+      "learning_rate": 2.9885411811226547e-05,
+      "loss": 1.0415,
+      "step": 95
+    },
+    {
+      "epoch": 0.4672897196261682,
+      "grad_norm": 0.5554174184799194,
+      "learning_rate": 2.9855023438887168e-05,
+      "loss": 1.0701,
+      "step": 100
+    },
+    {
+      "epoch": 0.49065420560747663,
+      "grad_norm": 0.495527058839798,
+      "learning_rate": 2.9821084341622924e-05,
+      "loss": 1.0041,
+      "step": 105
+    },
+    {
+      "epoch": 0.514018691588785,
+      "grad_norm": 0.5037195086479187,
+      "learning_rate": 2.9783602631733272e-05,
+      "loss": 1.0189,
+      "step": 110
+    },
+    {
+      "epoch": 0.5373831775700935,
+      "grad_norm": 0.5182456374168396,
+      "learning_rate": 2.974258726829136e-05,
+      "loss": 0.9733,
+      "step": 115
+    },
+    {
+      "epoch": 0.5607476635514018,
+      "grad_norm": 0.5680457949638367,
+      "learning_rate": 2.9698048055002566e-05,
+      "loss": 0.9747,
+      "step": 120
+    },
+    {
+      "epoch": 0.5841121495327103,
+      "grad_norm": 0.5237352252006531,
+      "learning_rate": 2.9649995637861182e-05,
+      "loss": 0.9813,
+      "step": 125
+    },
+    {
+      "epoch": 0.6074766355140186,
+      "grad_norm": 0.6517245769500732,
+      "learning_rate": 2.959844150260576e-05,
+      "loss": 0.8996,
+      "step": 130
+    },
+    {
+      "epoch": 0.6308411214953271,
+      "grad_norm": 0.624985933303833,
+      "learning_rate": 2.95433979719737e-05,
+      "loss": 0.9549,
+      "step": 135
+    },
+    {
+      "epoch": 0.6542056074766355,
+      "grad_norm": 0.721772313117981,
+      "learning_rate": 2.9484878202755834e-05,
+      "loss": 0.9164,
+      "step": 140
+    },
+    {
+      "epoch": 0.677570093457944,
+      "grad_norm": 0.6142625212669373,
+      "learning_rate": 2.942289618265161e-05,
+      "loss": 0.9402,
+      "step": 145
+    },
+    {
+      "epoch": 0.7009345794392523,
+      "grad_norm": 0.7766824960708618,
+      "learning_rate": 2.9357466726925683e-05,
+      "loss": 0.8848,
+      "step": 150
+    },
+    {
+      "epoch": 0.7242990654205608,
+      "grad_norm": 0.7013019323348999,
+      "learning_rate": 2.9288605474866687e-05,
+      "loss": 0.8473,
+      "step": 155
+    },
+    {
+      "epoch": 0.7476635514018691,
+      "grad_norm": 0.6742664575576782,
+      "learning_rate": 2.921632888604906e-05,
+      "loss": 0.87,
+      "step": 160
+    },
+    {
+      "epoch": 0.7710280373831776,
+      "grad_norm": 0.6647071838378906,
+      "learning_rate": 2.9140654236398786e-05,
+      "loss": 0.8952,
+      "step": 165
+    },
+    {
+      "epoch": 0.794392523364486,
+      "grad_norm": 0.8462722897529602,
+      "learning_rate": 2.906159961406401e-05,
+      "loss": 0.8788,
+      "step": 170
+    },
+    {
+      "epoch": 0.8177570093457944,
+      "grad_norm": 0.756624162197113,
+      "learning_rate": 2.8979183915091518e-05,
+      "loss": 0.8308,
+      "step": 175
+    },
+    {
+      "epoch": 0.8411214953271028,
+      "grad_norm": 0.7348244190216064,
+      "learning_rate": 2.8893426838910116e-05,
+      "loss": 0.8373,
+      "step": 180
+    },
+    {
+      "epoch": 0.8644859813084113,
+      "grad_norm": 0.7541269659996033,
+      "learning_rate": 2.8804348883621948e-05,
+      "loss": 0.8233,
+      "step": 185
+    },
+    {
+      "epoch": 0.8878504672897196,
+      "grad_norm": 0.8302907943725586,
+      "learning_rate": 2.8711971341102952e-05,
+      "loss": 0.7956,
+      "step": 190
+    },
+    {
+      "epoch": 0.9112149532710281,
+      "grad_norm": 0.7943097949028015,
+      "learning_rate": 2.861631629191356e-05,
+      "loss": 0.7513,
+      "step": 195
+    },
+    {
+      "epoch": 0.9345794392523364,
+      "grad_norm": 0.8179221749305725,
+      "learning_rate": 2.8517406600020876e-05,
+      "loss": 0.8219,
+      "step": 200
+    },
+    {
+      "epoch": 0.9579439252336449,
+      "grad_norm": 0.779038667678833,
+      "learning_rate": 2.841526590733364e-05,
+      "loss": 0.7927,
+      "step": 205
+    },
+    {
+      "epoch": 0.9813084112149533,
+      "grad_norm": 0.8207105994224548,
+      "learning_rate": 2.8309918628051177e-05,
+      "loss": 0.7336,
+      "step": 210
+    },
+    {
+      "epoch": 1.0046728971962617,
+      "grad_norm": 0.9244384765625,
+      "learning_rate": 2.8201389942827827e-05,
+      "loss": 0.7452,
+      "step": 215
+    },
+    {
+      "epoch": 1.02803738317757,
+      "grad_norm": 1.1412228345870972,
+      "learning_rate": 2.8089705792754107e-05,
+      "loss": 0.6897,
+      "step": 220
+    },
+    {
+      "epoch": 1.0514018691588785,
+      "grad_norm": 0.9332628846168518,
+      "learning_rate": 2.797489287315614e-05,
+      "loss": 0.6885,
+      "step": 225
+    },
+    {
+      "epoch": 1.074766355140187,
+      "grad_norm": 0.9464613199234009,
+      "learning_rate": 2.7856978627214824e-05,
+      "loss": 0.6619,
+      "step": 230
+    },
+    {
+      "epoch": 1.0981308411214954,
+      "grad_norm": 0.887984037399292,
+      "learning_rate": 2.7735991239406184e-05,
+      "loss": 0.6558,
+      "step": 235
+    },
+    {
+      "epoch": 1.1214953271028036,
+      "grad_norm": 1.1088521480560303,
+      "learning_rate": 2.7611959628764595e-05,
+      "loss": 0.6338,
+      "step": 240
+    },
+    {
+      "epoch": 1.144859813084112,
+      "grad_norm": 0.9827927947044373,
+      "learning_rate": 2.748491344197041e-05,
+      "loss": 0.6683,
+      "step": 245
+    },
+    {
+      "epoch": 1.1682242990654206,
+      "grad_norm": 0.9858003854751587,
+      "learning_rate": 2.735488304626362e-05,
+      "loss": 0.6383,
+      "step": 250
+    },
+    {
+      "epoch": 1.191588785046729,
+      "grad_norm": 0.9236980676651001,
+      "learning_rate": 2.7221899522185365e-05,
+      "loss": 0.6244,
+      "step": 255
+    },
+    {
+      "epoch": 1.2149532710280373,
+      "grad_norm": 0.9229886531829834,
+      "learning_rate": 2.708599465614888e-05,
+      "loss": 0.6332,
+      "step": 260
+    },
+    {
+      "epoch": 1.2383177570093458,
+      "grad_norm": 1.2654668092727661,
+      "learning_rate": 2.6947200932841723e-05,
+      "loss": 0.6041,
+      "step": 265
+    },
+    {
+      "epoch": 1.2616822429906542,
+      "grad_norm": 1.0240631103515625,
+      "learning_rate": 2.680555152746115e-05,
+      "loss": 0.6264,
+      "step": 270
+    },
+    {
+      "epoch": 1.2850467289719627,
+      "grad_norm": 1.0189911127090454,
+      "learning_rate": 2.6661080297784387e-05,
+      "loss": 0.6265,
+      "step": 275
+    },
+    {
+      "epoch": 1.308411214953271,
+      "grad_norm": 0.9291442632675171,
+      "learning_rate": 2.6513821776075772e-05,
+      "loss": 0.5748,
+      "step": 280
+    },
+    {
+      "epoch": 1.3317757009345794,
+      "grad_norm": 0.9921167492866516,
+      "learning_rate": 2.6363811160832678e-05,
+      "loss": 0.609,
+      "step": 285
+    },
+    {
+      "epoch": 1.355140186915888,
+      "grad_norm": 0.9764885306358337,
+      "learning_rate": 2.621108430837217e-05,
+      "loss": 0.5683,
+      "step": 290
+    },
+    {
+      "epoch": 1.3785046728971961,
+      "grad_norm": 1.087034821510315,
+      "learning_rate": 2.605567772426046e-05,
+      "loss": 0.5731,
+      "step": 295
+    },
+    {
+      "epoch": 1.4018691588785046,
+      "grad_norm": 0.966983437538147,
+      "learning_rate": 2.5897628554587152e-05,
+      "loss": 0.5679,
+      "step": 300
+    },
+    {
+      "epoch": 1.425233644859813,
+      "grad_norm": 0.9949087500572205,
+      "learning_rate": 2.573697457708639e-05,
+      "loss": 0.573,
+      "step": 305
+    },
+    {
+      "epoch": 1.4485981308411215,
+      "grad_norm": 1.0144646167755127,
+      "learning_rate": 2.5573754192107015e-05,
+      "loss": 0.5505,
+      "step": 310
+    },
+    {
+      "epoch": 1.47196261682243,
+      "grad_norm": 1.0582274198532104,
+      "learning_rate": 2.5408006413433937e-05,
+      "loss": 0.5335,
+      "step": 315
+    },
+    {
+      "epoch": 1.4953271028037383,
+      "grad_norm": 0.98384690284729,
+      "learning_rate": 2.5239770858962833e-05,
+      "loss": 0.6096,
+      "step": 320
+    },
+    {
+      "epoch": 1.5186915887850467,
+      "grad_norm": 0.9911113381385803,
+      "learning_rate": 2.5069087741230493e-05,
+      "loss": 0.5624,
+      "step": 325
+    },
+    {
+      "epoch": 1.542056074766355,
+      "grad_norm": 1.0836067199707031,
+      "learning_rate": 2.4895997857802997e-05,
+      "loss": 0.5683,
+      "step": 330
+    },
+    {
+      "epoch": 1.5654205607476634,
+      "grad_norm": 0.9900575876235962,
+      "learning_rate": 2.472054258152408e-05,
+      "loss": 0.549,
+      "step": 335
+    },
+    {
+      "epoch": 1.588785046728972,
+      "grad_norm": 1.1659897565841675,
+      "learning_rate": 2.454276385062596e-05,
+      "loss": 0.5428,
+      "step": 340
+    },
+    {
+      "epoch": 1.6121495327102804,
+      "grad_norm": 1.314231514930725,
+      "learning_rate": 2.4362704158705064e-05,
+      "loss": 0.5119,
+      "step": 345
+    },
+    {
+      "epoch": 1.6355140186915889,
+      "grad_norm": 0.9947055578231812,
+      "learning_rate": 2.4180406544564983e-05,
+      "loss": 0.5302,
+      "step": 350
+    },
+    {
+      "epoch": 1.6588785046728973,
+      "grad_norm": 1.0473575592041016,
+      "learning_rate": 2.39959145819291e-05,
+      "loss": 0.5321,
+      "step": 355
+    },
+    {
+      "epoch": 1.6822429906542056,
+      "grad_norm": 1.1833536624908447,
+      "learning_rate": 2.380927236902539e-05,
+      "loss": 0.5358,
+      "step": 360
+    },
+    {
+      "epoch": 1.705607476635514,
+      "grad_norm": 1.179697036743164,
+      "learning_rate": 2.3620524518045826e-05,
+      "loss": 0.507,
+      "step": 365
+    },
+    {
+      "epoch": 1.7289719626168223,
+      "grad_norm": 1.0846567153930664,
+      "learning_rate": 2.3429716144482947e-05,
+      "loss": 0.501,
+      "step": 370
+    },
+    {
+      "epoch": 1.7523364485981308,
+      "grad_norm": 1.0222374200820923,
+      "learning_rate": 2.323689285634609e-05,
+      "loss": 0.5085,
+      "step": 375
+    },
+    {
+      "epoch": 1.7757009345794392,
+      "grad_norm": 1.3095283508300781,
+      "learning_rate": 2.304210074325996e-05,
+      "loss": 0.4818,
+      "step": 380
+    },
+    {
+      "epoch": 1.7990654205607477,
+      "grad_norm": 1.1077048778533936,
+      "learning_rate": 2.2845386365448043e-05,
+      "loss": 0.4877,
+      "step": 385
+    },
+    {
+      "epoch": 1.8224299065420562,
+      "grad_norm": 1.181553602218628,
+      "learning_rate": 2.2646796742603516e-05,
+      "loss": 0.4752,
+      "step": 390
+    },
+    {
+      "epoch": 1.8457943925233646,
+      "grad_norm": 0.9620625972747803,
+      "learning_rate": 2.2446379342650396e-05,
+      "loss": 0.4793,
+      "step": 395
+    },
+    {
+      "epoch": 1.8691588785046729,
+      "grad_norm": 1.1335831880569458,
+      "learning_rate": 2.224418207039746e-05,
+      "loss": 0.488,
+      "step": 400
+    },
+    {
+      "epoch": 1.8925233644859814,
+      "grad_norm": 1.0170129537582397,
+      "learning_rate": 2.204025325608783e-05,
+      "loss": 0.4817,
+      "step": 405
+    },
+    {
+      "epoch": 1.9158878504672896,
+      "grad_norm": 1.1246532201766968,
+      "learning_rate": 2.1834641643846807e-05,
+      "loss": 0.4863,
+      "step": 410
+    },
+    {
+      "epoch": 1.939252336448598,
+      "grad_norm": 1.0550944805145264,
+      "learning_rate": 2.1627396380030803e-05,
+      "loss": 0.49,
+      "step": 415
+    },
+    {
+      "epoch": 1.9626168224299065,
+      "grad_norm": 1.0045976638793945,
+      "learning_rate": 2.141856700148012e-05,
+      "loss": 0.4585,
+      "step": 420
+    },
+    {
+      "epoch": 1.985981308411215,
+      "grad_norm": 1.1983375549316406,
+      "learning_rate": 2.1208203423678402e-05,
+      "loss": 0.4538,
+      "step": 425
+    },
+    {
+      "epoch": 2.0093457943925235,
+      "grad_norm": 1.2544827461242676,
+      "learning_rate": 2.099635592882157e-05,
+      "loss": 0.4146,
+      "step": 430
+    },
+    {
+      "epoch": 2.032710280373832,
+      "grad_norm": 1.2281593084335327,
+      "learning_rate": 2.0783075153799112e-05,
+      "loss": 0.375,
+      "step": 435
+    },
+    {
+      "epoch": 2.05607476635514,
+      "grad_norm": 1.1033180952072144,
+      "learning_rate": 2.0568412078090597e-05,
+      "loss": 0.3817,
+      "step": 440
+    },
+    {
+      "epoch": 2.0794392523364484,
+      "grad_norm": 1.1841470003128052,
+      "learning_rate": 2.0352418011580293e-05,
+      "loss": 0.3811,
+      "step": 445
+    },
+    {
+      "epoch": 2.102803738317757,
+      "grad_norm": 1.223242998123169,
+      "learning_rate": 2.0135144582292796e-05,
+      "loss": 0.3673,
+      "step": 450
+    },
+    {
+      "epoch": 2.1261682242990654,
+      "grad_norm": 1.1610691547393799,
+      "learning_rate": 1.991664372405265e-05,
+      "loss": 0.3813,
+      "step": 455
+    },
+    {
+      "epoch": 2.149532710280374,
+      "grad_norm": 1.187243938446045,
+      "learning_rate": 1.9696967664070854e-05,
+      "loss": 0.3899,
+      "step": 460
+    },
+    {
+      "epoch": 2.1728971962616823,
+      "grad_norm": 1.1279627084732056,
+      "learning_rate": 1.947616891046123e-05,
+      "loss": 0.3744,
+      "step": 465
+    },
+    {
+      "epoch": 2.196261682242991,
+      "grad_norm": 1.23040771484375,
+      "learning_rate": 1.925430023968967e-05,
+      "loss": 0.3773,
+      "step": 470
+    },
+    {
+      "epoch": 2.2196261682242993,
+      "grad_norm": 1.1494715213775635,
+      "learning_rate": 1.9031414683959214e-05,
+      "loss": 0.3782,
+      "step": 475
+    },
+    {
+      "epoch": 2.2429906542056073,
+      "grad_norm": 1.3942996263504028,
+      "learning_rate": 1.8807565518534055e-05,
+      "loss": 0.3604,
+      "step": 480
+    },
+    {
+      "epoch": 2.2663551401869158,
+      "grad_norm": 1.135150671005249,
+      "learning_rate": 1.858280624900537e-05,
+      "loss": 0.3621,
+      "step": 485
+    },
+    {
+      "epoch": 2.289719626168224,
+      "grad_norm": 1.2644059658050537,
+      "learning_rate": 1.835719059850215e-05,
+      "loss": 0.3579,
+      "step": 490
+    },
+    {
+      "epoch": 2.3130841121495327,
+      "grad_norm": 1.16321861743927,
+      "learning_rate": 1.813077249485003e-05,
+      "loss": 0.3781,
+      "step": 495
+    },
+    {
+      "epoch": 2.336448598130841,
+      "grad_norm": 1.1858952045440674,
+      "learning_rate": 1.7903606057681143e-05,
+      "loss": 0.3489,
+      "step": 500
+    },
+    {
+      "epoch": 2.3598130841121496,
+      "grad_norm": 1.2584055662155151,
+      "learning_rate": 1.7675745585498224e-05,
+      "loss": 0.3238,
+      "step": 505
+    },
+    {
+      "epoch": 2.383177570093458,
+      "grad_norm": 1.1378344297409058,
+      "learning_rate": 1.744724554269583e-05,
+      "loss": 0.3632,
+      "step": 510
+    },
+    {
+      "epoch": 2.406542056074766,
+      "grad_norm": 1.1574376821517944,
+      "learning_rate": 1.7218160546542037e-05,
+      "loss": 0.3151,
+      "step": 515
+    },
+    {
+      "epoch": 2.4299065420560746,
+      "grad_norm": 1.0761138200759888,
+      "learning_rate": 1.69885453541235e-05,
+      "loss": 0.327,
+      "step": 520
+    },
+    {
+      "epoch": 2.453271028037383,
+      "grad_norm": 1.1991745233535767,
+      "learning_rate": 1.6758454849257147e-05,
+      "loss": 0.3239,
+      "step": 525
+    },
+    {
+      "epoch": 2.4766355140186915,
+      "grad_norm": 1.3082544803619385,
+      "learning_rate": 1.652794402937156e-05,
+      "loss": 0.3149,
+      "step": 530
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.2072352170944214,
+      "learning_rate": 1.6297067992361185e-05,
+      "loss": 0.3252,
+      "step": 535
+    },
+    {
+      "epoch": 2.5233644859813085,
+      "grad_norm": 1.217516303062439,
+      "learning_rate": 1.6065881923416592e-05,
+      "loss": 0.3288,
+      "step": 540
+    },
+    {
+      "epoch": 2.546728971962617,
+      "grad_norm": 1.1310899257659912,
+      "learning_rate": 1.5834441081833768e-05,
+      "loss": 0.3466,
+      "step": 545
+    },
+    {
+      "epoch": 2.5700934579439254,
+      "grad_norm": 1.1097090244293213,
+      "learning_rate": 1.56028007878058e-05,
+      "loss": 0.3285,
+      "step": 550
+    },
+    {
+      "epoch": 2.593457943925234,
+      "grad_norm": 1.1151643991470337,
+      "learning_rate": 1.5371016409199926e-05,
+      "loss": 0.3005,
+      "step": 555
+    },
+    {
+      "epoch": 2.616822429906542,
+      "grad_norm": 1.2807705402374268,
+      "learning_rate": 1.5139143348323227e-05,
+      "loss": 0.3094,
+      "step": 560
+    },
+    {
+      "epoch": 2.6401869158878504,
+      "grad_norm": 1.1719436645507812,
+      "learning_rate": 1.4907237028680091e-05,
+      "loss": 0.3091,
+      "step": 565
+    },
+    {
+      "epoch": 2.663551401869159,
+      "grad_norm": 1.2889381647109985,
+      "learning_rate": 1.4675352881724576e-05,
+      "loss": 0.309,
+      "step": 570
+    },
+    {
+      "epoch": 2.6869158878504673,
+      "grad_norm": 1.2569345235824585,
+      "learning_rate": 1.4443546333610923e-05,
+      "loss": 0.2934,
+      "step": 575
+    },
+    {
+      "epoch": 2.710280373831776,
+      "grad_norm": 1.1071770191192627,
+      "learning_rate": 1.4211872791945282e-05,
+      "loss": 0.297,
+      "step": 580
+    },
+    {
+      "epoch": 2.7336448598130842,
+      "grad_norm": 1.2480145692825317,
+      "learning_rate": 1.3980387632541923e-05,
+      "loss": 0.3076,
+      "step": 585
+    },
+    {
+      "epoch": 2.7570093457943923,
+      "grad_norm": 1.2675397396087646,
+      "learning_rate": 1.374914618618699e-05,
+      "loss": 0.3112,
+      "step": 590
+    },
+    {
+      "epoch": 2.7803738317757007,
+      "grad_norm": 1.1788510084152222,
+      "learning_rate": 1.3518203725413086e-05,
+      "loss": 0.3028,
+      "step": 595
+    },
+    {
+      "epoch": 2.803738317757009,
+      "grad_norm": 1.194700002670288,
+      "learning_rate": 1.3287615451287722e-05,
+      "loss": 0.3172,
+      "step": 600
+    },
+    {
+      "epoch": 2.8271028037383177,
+      "grad_norm": 1.2842059135437012,
+      "learning_rate": 1.3057436480218866e-05,
+      "loss": 0.308,
+      "step": 605
+    },
+    {
+      "epoch": 2.850467289719626,
+      "grad_norm": 1.28480863571167,
+      "learning_rate": 1.2827721830780755e-05,
+      "loss": 0.2993,
+      "step": 610
+    },
+    {
+      "epoch": 2.8738317757009346,
+      "grad_norm": 1.30231511592865,
+      "learning_rate": 1.2598526410563037e-05,
+      "loss": 0.2897,
+      "step": 615
+    },
+    {
+      "epoch": 2.897196261682243,
+      "grad_norm": 1.3034412860870361,
+      "learning_rate": 1.2369905003046514e-05,
+      "loss": 0.2863,
+      "step": 620
+    },
+    {
+      "epoch": 2.9205607476635516,
+      "grad_norm": 1.1544686555862427,
+      "learning_rate": 1.2141912254508477e-05,
+      "loss": 0.2871,
+      "step": 625
+    },
+    {
+      "epoch": 2.94392523364486,
+      "grad_norm": 1.2238121032714844,
+      "learning_rate": 1.1914602660960873e-05,
+      "loss": 0.2781,
+      "step": 630
+    },
+    {
+      "epoch": 2.9672897196261685,
+      "grad_norm": 1.3411394357681274,
+      "learning_rate": 1.1688030555124409e-05,
+      "loss": 0.2776,
+      "step": 635
+    },
+    {
+      "epoch": 2.9906542056074765,
+      "grad_norm": 1.2078008651733398,
+      "learning_rate": 1.1462250093441632e-05,
+      "loss": 0.2511,
+      "step": 640
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1070,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4791784940290703e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97291ef091bf25732b2bcf43f778fd55f897d7eb6fb98a7aaa7f6d4f5b994519
+size 7800

hotpotqa/not_grounded/1_128_e5_3e-5/checkpoint-642/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)

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/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 0.12.0

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "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_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": [
+    "k_proj",
+    "v_proj",
+    "q_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd82eaa7feed84c14a6210930bd581d2e7a36dd96b50c55f0c3f816ba812f573
+size 791751704

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/latest

@@ -0,0 +1 @@
+global_step682

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b057049dd3265ea3e5772b8b789fac5cdde9997724c3886773877506365eef21
+size 15920

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed24384ec43550c3b40277b675faa7da3106c41739b5e487403bdfe4ef6ad22c
+size 15920

hotpotqa/not_grounded/2_128_e5_3e-5/checkpoint-684/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98297d4c36b94b26b02d1e833595bf71dc3faf52ae85486c90224de1e19ef9b0
+size 15920