Upload folder using huggingface_hub

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/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/base/4_128_e5_3e-5/checkpoint-1060/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",
+    "up_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8b893b10a581ff12e8f44726c364d5f703a0dfa2d03b9cae333013500cdfbed
+size 791751704

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/latest

@@ -0,0 +1 @@
+global_step1058

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b952c6f8993a492bb83ef74c42dd7baf4fe13d370f683ccf8f91d388b4859730
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e3713541c65974dbf8fce44c359785cd0cdef65d7a29140d3ea201827fea60b
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:909d468064c7e309a164ac22ff6b287c2f5e3aafbc9241304d1aa2ec510f3584
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d14d48dba05234045fbf9c742daf8af21780bf885f1492881e2de311b794419e
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8783ece37c963ca4ed5f2336925564b56bb96499544f40592723e9d306fc67c0
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66dbb60bfc2366a5e7e176933826c31f4514df77530acb6ac582cf69ffed20e3
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6877f7e15e640acad32e15ed7d597d22b11db8d7d21070bc34c2fd0fc784e5e3
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5332764cc5b0c540daae2fa604fa898abaab5697282cface83ff996c330412c8
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a8bcf4c11e7a0e39f424d34d17ecc8dfc8d74218eb57752dd51ffd1c6ab4880
+size 1064

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/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/base/4_128_e5_3e-5/checkpoint-1060/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/base/4_128_e5_3e-5/checkpoint-1060/trainer_state.json

@@ -0,0 +1,1518 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.978823529411764,
+  "eval_steps": 500,
+  "global_step": 1060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.023529411764705882,
+      "grad_norm": 0.8258791565895081,
+      "learning_rate": 2.2641509433962262e-06,
+      "loss": 1.3394,
+      "step": 5
+    },
+    {
+      "epoch": 0.047058823529411764,
+      "grad_norm": 0.5909937024116516,
+      "learning_rate": 5.094339622641509e-06,
+      "loss": 1.3138,
+      "step": 10
+    },
+    {
+      "epoch": 0.07058823529411765,
+      "grad_norm": 0.5355730652809143,
+      "learning_rate": 7.924528301886793e-06,
+      "loss": 1.3315,
+      "step": 15
+    },
+    {
+      "epoch": 0.09411764705882353,
+      "grad_norm": 0.5410304665565491,
+      "learning_rate": 1.0754716981132076e-05,
+      "loss": 1.3214,
+      "step": 20
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 0.4087582230567932,
+      "learning_rate": 1.358490566037736e-05,
+      "loss": 1.2642,
+      "step": 25
+    },
+    {
+      "epoch": 0.1411764705882353,
+      "grad_norm": 0.3983219861984253,
+      "learning_rate": 1.6415094339622643e-05,
+      "loss": 1.2698,
+      "step": 30
+    },
+    {
+      "epoch": 0.16470588235294117,
+      "grad_norm": 0.39306771755218506,
+      "learning_rate": 1.9245283018867924e-05,
+      "loss": 1.2605,
+      "step": 35
+    },
+    {
+      "epoch": 0.18823529411764706,
+      "grad_norm": 0.3943307101726532,
+      "learning_rate": 2.2075471698113208e-05,
+      "loss": 1.2468,
+      "step": 40
+    },
+    {
+      "epoch": 0.21176470588235294,
+      "grad_norm": 0.34280142188072205,
+      "learning_rate": 2.4905660377358492e-05,
+      "loss": 1.2487,
+      "step": 45
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 0.39148157835006714,
+      "learning_rate": 2.7735849056603773e-05,
+      "loss": 1.2446,
+      "step": 50
+    },
+    {
+      "epoch": 0.25882352941176473,
+      "grad_norm": 0.44964322447776794,
+      "learning_rate": 2.99999270035541e-05,
+      "loss": 1.2281,
+      "step": 55
+    },
+    {
+      "epoch": 0.2823529411764706,
+      "grad_norm": 0.40044668316841125,
+      "learning_rate": 2.999737220254547e-05,
+      "loss": 1.1738,
+      "step": 60
+    },
+    {
+      "epoch": 0.3058823529411765,
+      "grad_norm": 0.45404553413391113,
+      "learning_rate": 2.999116828967479e-05,
+      "loss": 1.1659,
+      "step": 65
+    },
+    {
+      "epoch": 0.32941176470588235,
+      "grad_norm": 0.4405958652496338,
+      "learning_rate": 2.9981316774457966e-05,
+      "loss": 1.1103,
+      "step": 70
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 0.43754836916923523,
+      "learning_rate": 2.9967820053933663e-05,
+      "loss": 1.1569,
+      "step": 75
+    },
+    {
+      "epoch": 0.3764705882352941,
+      "grad_norm": 0.48351433873176575,
+      "learning_rate": 2.995068141208006e-05,
+      "loss": 1.0878,
+      "step": 80
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.45371121168136597,
+      "learning_rate": 2.992990501901579e-05,
+      "loss": 1.0451,
+      "step": 85
+    },
+    {
+      "epoch": 0.4235294117647059,
+      "grad_norm": 0.4767340123653412,
+      "learning_rate": 2.9905495929985282e-05,
+      "loss": 1.0701,
+      "step": 90
+    },
+    {
+      "epoch": 0.4470588235294118,
+      "grad_norm": 0.6476609110832214,
+      "learning_rate": 2.987746008412876e-05,
+      "loss": 1.0156,
+      "step": 95
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 0.4992612302303314,
+      "learning_rate": 2.984580430303711e-05,
+      "loss": 1.0879,
+      "step": 100
+    },
+    {
+      "epoch": 0.49411764705882355,
+      "grad_norm": 0.5524128079414368,
+      "learning_rate": 2.98105362890921e-05,
+      "loss": 1.0219,
+      "step": 105
+    },
+    {
+      "epoch": 0.5176470588235295,
+      "grad_norm": 0.7320187091827393,
+      "learning_rate": 2.9771664623592276e-05,
+      "loss": 1.0294,
+      "step": 110
+    },
+    {
+      "epoch": 0.5411764705882353,
+      "grad_norm": 0.5769292116165161,
+      "learning_rate": 2.9729198764664944e-05,
+      "loss": 0.9888,
+      "step": 115
+    },
+    {
+      "epoch": 0.5647058823529412,
+      "grad_norm": 0.6180774569511414,
+      "learning_rate": 2.968314904496488e-05,
+      "loss": 1.0136,
+      "step": 120
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 0.6959433555603027,
+      "learning_rate": 2.9633526669160217e-05,
+      "loss": 1.0116,
+      "step": 125
+    },
+    {
+      "epoch": 0.611764705882353,
+      "grad_norm": 0.7187851071357727,
+      "learning_rate": 2.9580343711206162e-05,
+      "loss": 0.9455,
+      "step": 130
+    },
+    {
+      "epoch": 0.6352941176470588,
+      "grad_norm": 0.7223728895187378,
+      "learning_rate": 2.9523613111407196e-05,
+      "loss": 0.9332,
+      "step": 135
+    },
+    {
+      "epoch": 0.6588235294117647,
+      "grad_norm": 0.6712695956230164,
+      "learning_rate": 2.946334867326849e-05,
+      "loss": 0.9047,
+      "step": 140
+    },
+    {
+      "epoch": 0.6823529411764706,
+      "grad_norm": 0.6632208824157715,
+      "learning_rate": 2.939956506013727e-05,
+      "loss": 0.9201,
+      "step": 145
+    },
+    {
+      "epoch": 0.7058823529411765,
+      "grad_norm": 0.7727546691894531,
+      "learning_rate": 2.933227779163497e-05,
+      "loss": 0.8752,
+      "step": 150
+    },
+    {
+      "epoch": 0.7294117647058823,
+      "grad_norm": 0.7360590100288391,
+      "learning_rate": 2.926150323988108e-05,
+      "loss": 0.8369,
+      "step": 155
+    },
+    {
+      "epoch": 0.7529411764705882,
+      "grad_norm": 0.7336561679840088,
+      "learning_rate": 2.9187258625509518e-05,
+      "loss": 0.8651,
+      "step": 160
+    },
+    {
+      "epoch": 0.7764705882352941,
+      "grad_norm": 0.7670441269874573,
+      "learning_rate": 2.9109562013478553e-05,
+      "loss": 0.8366,
+      "step": 165
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.7336223721504211,
+      "learning_rate": 2.9028432308675305e-05,
+      "loss": 0.9056,
+      "step": 170
+    },
+    {
+      "epoch": 0.8235294117647058,
+      "grad_norm": 0.8804784417152405,
+      "learning_rate": 2.8943889251315867e-05,
+      "loss": 0.8196,
+      "step": 175
+    },
+    {
+      "epoch": 0.8470588235294118,
+      "grad_norm": 0.8821813464164734,
+      "learning_rate": 2.8855953412142184e-05,
+      "loss": 0.7661,
+      "step": 180
+    },
+    {
+      "epoch": 0.8705882352941177,
+      "grad_norm": 1.065437912940979,
+      "learning_rate": 2.876464618741686e-05,
+      "loss": 0.8047,
+      "step": 185
+    },
+    {
+      "epoch": 0.8941176470588236,
+      "grad_norm": 0.957807183265686,
+      "learning_rate": 2.866998979371705e-05,
+      "loss": 0.8138,
+      "step": 190
+    },
+    {
+      "epoch": 0.9176470588235294,
+      "grad_norm": 0.8701325058937073,
+      "learning_rate": 2.857200726252883e-05,
+      "loss": 0.7415,
+      "step": 195
+    },
+    {
+      "epoch": 0.9411764705882353,
+      "grad_norm": 0.8715978860855103,
+      "learning_rate": 2.847072243464325e-05,
+      "loss": 0.8062,
+      "step": 200
+    },
+    {
+      "epoch": 0.9647058823529412,
+      "grad_norm": 0.8279728889465332,
+      "learning_rate": 2.8366159954355444e-05,
+      "loss": 0.7496,
+      "step": 205
+    },
+    {
+      "epoch": 0.9882352941176471,
+      "grad_norm": 0.8706197142601013,
+      "learning_rate": 2.8258345263468284e-05,
+      "loss": 0.721,
+      "step": 210
+    },
+    {
+      "epoch": 1.0094117647058825,
+      "grad_norm": 0.9178848266601562,
+      "learning_rate": 2.8147304595101952e-05,
+      "loss": 0.7232,
+      "step": 215
+    },
+    {
+      "epoch": 1.0329411764705883,
+      "grad_norm": 0.9863954186439514,
+      "learning_rate": 2.8033064967310985e-05,
+      "loss": 0.6979,
+      "step": 220
+    },
+    {
+      "epoch": 1.056470588235294,
+      "grad_norm": 1.0150811672210693,
+      "learning_rate": 2.791565417651033e-05,
+      "loss": 0.6858,
+      "step": 225
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 1.0401983261108398,
+      "learning_rate": 2.7795100790712027e-05,
+      "loss": 0.674,
+      "step": 230
+    },
+    {
+      "epoch": 1.1035294117647059,
+      "grad_norm": 1.0883874893188477,
+      "learning_rate": 2.7671434142574084e-05,
+      "loss": 0.6526,
+      "step": 235
+    },
+    {
+      "epoch": 1.1270588235294117,
+      "grad_norm": 1.0119273662567139,
+      "learning_rate": 2.7544684322263416e-05,
+      "loss": 0.6245,
+      "step": 240
+    },
+    {
+      "epoch": 1.1505882352941177,
+      "grad_norm": 1.0836514234542847,
+      "learning_rate": 2.7414882170134355e-05,
+      "loss": 0.6019,
+      "step": 245
+    },
+    {
+      "epoch": 1.1741176470588235,
+      "grad_norm": 1.0072085857391357,
+      "learning_rate": 2.7282059269224688e-05,
+      "loss": 0.6535,
+      "step": 250
+    },
+    {
+      "epoch": 1.1976470588235295,
+      "grad_norm": 1.060310959815979,
+      "learning_rate": 2.714624793757099e-05,
+      "loss": 0.6506,
+      "step": 255
+    },
+    {
+      "epoch": 1.2211764705882353,
+      "grad_norm": 1.047108769416809,
+      "learning_rate": 2.7007481220345077e-05,
+      "loss": 0.631,
+      "step": 260
+    },
+    {
+      "epoch": 1.244705882352941,
+      "grad_norm": 0.9935978651046753,
+      "learning_rate": 2.686579288181359e-05,
+      "loss": 0.6502,
+      "step": 265
+    },
+    {
+      "epoch": 1.2682352941176471,
+      "grad_norm": 1.187946081161499,
+      "learning_rate": 2.6721217397122565e-05,
+      "loss": 0.6152,
+      "step": 270
+    },
+    {
+      "epoch": 1.291764705882353,
+      "grad_norm": 1.918660283088684,
+      "learning_rate": 2.6573789943909064e-05,
+      "loss": 0.5633,
+      "step": 275
+    },
+    {
+      "epoch": 1.315294117647059,
+      "grad_norm": 1.0726563930511475,
+      "learning_rate": 2.6423546393741844e-05,
+      "loss": 0.6145,
+      "step": 280
+    },
+    {
+      "epoch": 1.3388235294117647,
+      "grad_norm": 1.0543569326400757,
+      "learning_rate": 2.6270523303393244e-05,
+      "loss": 0.5726,
+      "step": 285
+    },
+    {
+      "epoch": 1.3623529411764705,
+      "grad_norm": 1.0107314586639404,
+      "learning_rate": 2.6114757905944263e-05,
+      "loss": 0.5788,
+      "step": 290
+    },
+    {
+      "epoch": 1.3858823529411763,
+      "grad_norm": 1.0326977968215942,
+      "learning_rate": 2.5956288101725172e-05,
+      "loss": 0.535,
+      "step": 295
+    },
+    {
+      "epoch": 1.4094117647058824,
+      "grad_norm": 1.0717417001724243,
+      "learning_rate": 2.5795152449093696e-05,
+      "loss": 0.5982,
+      "step": 300
+    },
+    {
+      "epoch": 1.4329411764705882,
+      "grad_norm": 1.088898777961731,
+      "learning_rate": 2.5631390155053127e-05,
+      "loss": 0.5944,
+      "step": 305
+    },
+    {
+      "epoch": 1.4564705882352942,
+      "grad_norm": 0.9524811506271362,
+      "learning_rate": 2.5465041065712623e-05,
+      "loss": 0.5544,
+      "step": 310
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 1.1497600078582764,
+      "learning_rate": 2.529614565659196e-05,
+      "loss": 0.5633,
+      "step": 315
+    },
+    {
+      "epoch": 1.5035294117647058,
+      "grad_norm": 0.9918271899223328,
+      "learning_rate": 2.512474502277316e-05,
+      "loss": 0.5912,
+      "step": 320
+    },
+    {
+      "epoch": 1.5270588235294118,
+      "grad_norm": 1.0430793762207031,
+      "learning_rate": 2.4950880868901384e-05,
+      "loss": 0.5531,
+      "step": 325
+    },
+    {
+      "epoch": 1.5505882352941176,
+      "grad_norm": 1.1337051391601562,
+      "learning_rate": 2.4774595499037485e-05,
+      "loss": 0.5342,
+      "step": 330
+    },
+    {
+      "epoch": 1.5741176470588236,
+      "grad_norm": 1.180948257446289,
+      "learning_rate": 2.4595931806364718e-05,
+      "loss": 0.5162,
+      "step": 335
+    },
+    {
+      "epoch": 1.5976470588235294,
+      "grad_norm": 1.0531307458877563,
+      "learning_rate": 2.441493326275212e-05,
+      "loss": 0.5485,
+      "step": 340
+    },
+    {
+      "epoch": 1.6211764705882352,
+      "grad_norm": 1.0936872959136963,
+      "learning_rate": 2.4231643908177075e-05,
+      "loss": 0.5166,
+      "step": 345
+    },
+    {
+      "epoch": 1.644705882352941,
+      "grad_norm": 1.24155592918396,
+      "learning_rate": 2.404610834000963e-05,
+      "loss": 0.4678,
+      "step": 350
+    },
+    {
+      "epoch": 1.668235294117647,
+      "grad_norm": 1.2460212707519531,
+      "learning_rate": 2.385837170216123e-05,
+      "loss": 0.4981,
+      "step": 355
+    },
+    {
+      "epoch": 1.691764705882353,
+      "grad_norm": 0.9930285811424255,
+      "learning_rate": 2.366847967410044e-05,
+      "loss": 0.5294,
+      "step": 360
+    },
+    {
+      "epoch": 1.7152941176470589,
+      "grad_norm": 0.9697048664093018,
+      "learning_rate": 2.3476478459738375e-05,
+      "loss": 0.4993,
+      "step": 365
+    },
+    {
+      "epoch": 1.7388235294117647,
+      "grad_norm": 1.0432662963867188,
+      "learning_rate": 2.3282414776186506e-05,
+      "loss": 0.5097,
+      "step": 370
+    },
+    {
+      "epoch": 1.7623529411764705,
+      "grad_norm": 1.0439497232437134,
+      "learning_rate": 2.308633584238958e-05,
+      "loss": 0.4915,
+      "step": 375
+    },
+    {
+      "epoch": 1.7858823529411765,
+      "grad_norm": 1.2071393728256226,
+      "learning_rate": 2.288828936763651e-05,
+      "loss": 0.4804,
+      "step": 380
+    },
+    {
+      "epoch": 1.8094117647058825,
+      "grad_norm": 1.0262537002563477,
+      "learning_rate": 2.2688323539951876e-05,
+      "loss": 0.4604,
+      "step": 385
+    },
+    {
+      "epoch": 1.8329411764705883,
+      "grad_norm": 1.0664896965026855,
+      "learning_rate": 2.248648701437096e-05,
+      "loss": 0.484,
+      "step": 390
+    },
+    {
+      "epoch": 1.856470588235294,
+      "grad_norm": 1.029360294342041,
+      "learning_rate": 2.2282828901101194e-05,
+      "loss": 0.441,
+      "step": 395
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 1.077533483505249,
+      "learning_rate": 2.2077398753572803e-05,
+      "loss": 0.4381,
+      "step": 400
+    },
+    {
+      "epoch": 1.903529411764706,
+      "grad_norm": 1.0366874933242798,
+      "learning_rate": 2.1870246556381616e-05,
+      "loss": 0.473,
+      "step": 405
+    },
+    {
+      "epoch": 1.9270588235294117,
+      "grad_norm": 1.0753458738327026,
+      "learning_rate": 2.1661422713127002e-05,
+      "loss": 0.4518,
+      "step": 410
+    },
+    {
+      "epoch": 1.9505882352941177,
+      "grad_norm": 1.061562418937683,
+      "learning_rate": 2.145097803414781e-05,
+      "loss": 0.4652,
+      "step": 415
+    },
+    {
+      "epoch": 1.9741176470588235,
+      "grad_norm": 1.1651426553726196,
+      "learning_rate": 2.1238963724159376e-05,
+      "loss": 0.4638,
+      "step": 420
+    },
+    {
+      "epoch": 1.9976470588235293,
+      "grad_norm": 1.120908260345459,
+      "learning_rate": 2.1025431369794546e-05,
+      "loss": 0.4908,
+      "step": 425
+    },
+    {
+      "epoch": 2.018823529411765,
+      "grad_norm": 1.394509196281433,
+      "learning_rate": 2.081043292705179e-05,
+      "loss": 0.4035,
+      "step": 430
+    },
+    {
+      "epoch": 2.0423529411764707,
+      "grad_norm": 1.015360951423645,
+      "learning_rate": 2.059402070865345e-05,
+      "loss": 0.3415,
+      "step": 435
+    },
+    {
+      "epoch": 2.0658823529411765,
+      "grad_norm": 1.4450172185897827,
+      "learning_rate": 2.037624737131714e-05,
+      "loss": 0.3755,
+      "step": 440
+    },
+    {
+      "epoch": 2.0894117647058823,
+      "grad_norm": 1.201728343963623,
+      "learning_rate": 2.015716590294354e-05,
+      "loss": 0.3686,
+      "step": 445
+    },
+    {
+      "epoch": 2.112941176470588,
+      "grad_norm": 1.269172191619873,
+      "learning_rate": 1.9936829609723505e-05,
+      "loss": 0.3252,
+      "step": 450
+    },
+    {
+      "epoch": 2.1364705882352943,
+      "grad_norm": 1.1521567106246948,
+      "learning_rate": 1.971529210316779e-05,
+      "loss": 0.3818,
+      "step": 455
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 1.270796298980713,
+      "learning_rate": 1.9492607287062488e-05,
+      "loss": 0.3542,
+      "step": 460
+    },
+    {
+      "epoch": 2.183529411764706,
+      "grad_norm": 1.2757797241210938,
+      "learning_rate": 1.926882934435331e-05,
+      "loss": 0.379,
+      "step": 465
+    },
+    {
+      "epoch": 2.2070588235294117,
+      "grad_norm": 1.1761181354522705,
+      "learning_rate": 1.9044012723961986e-05,
+      "loss": 0.3625,
+      "step": 470
+    },
+    {
+      "epoch": 2.2305882352941175,
+      "grad_norm": 1.1094317436218262,
+      "learning_rate": 1.881821212753793e-05,
+      "loss": 0.3719,
+      "step": 475
+    },
+    {
+      "epoch": 2.2541176470588233,
+      "grad_norm": 1.241894006729126,
+      "learning_rate": 1.8591482496148397e-05,
+      "loss": 0.3533,
+      "step": 480
+    },
+    {
+      "epoch": 2.2776470588235296,
+      "grad_norm": 1.3358162641525269,
+      "learning_rate": 1.836387899691041e-05,
+      "loss": 0.3384,
+      "step": 485
+    },
+    {
+      "epoch": 2.3011764705882354,
+      "grad_norm": 1.1392518281936646,
+      "learning_rate": 1.8135457009567683e-05,
+      "loss": 0.3557,
+      "step": 490
+    },
+    {
+      "epoch": 2.324705882352941,
+      "grad_norm": 1.1473363637924194,
+      "learning_rate": 1.7906272113015793e-05,
+      "loss": 0.3687,
+      "step": 495
+    },
+    {
+      "epoch": 2.348235294117647,
+      "grad_norm": 1.249038815498352,
+      "learning_rate": 1.7676380071778904e-05,
+      "loss": 0.3461,
+      "step": 500
+    },
+    {
+      "epoch": 2.3717647058823528,
+      "grad_norm": 1.1381361484527588,
+      "learning_rate": 1.7445836822441328e-05,
+      "loss": 0.3559,
+      "step": 505
+    },
+    {
+      "epoch": 2.395294117647059,
+      "grad_norm": 1.2378185987472534,
+      "learning_rate": 1.721469846003722e-05,
+      "loss": 0.3247,
+      "step": 510
+    },
+    {
+      "epoch": 2.418823529411765,
+      "grad_norm": 1.218033790588379,
+      "learning_rate": 1.6983021224401705e-05,
+      "loss": 0.3445,
+      "step": 515
+    },
+    {
+      "epoch": 2.4423529411764706,
+      "grad_norm": 1.2067508697509766,
+      "learning_rate": 1.675086148648679e-05,
+      "loss": 0.3555,
+      "step": 520
+    },
+    {
+      "epoch": 2.4658823529411764,
+      "grad_norm": 1.1700276136398315,
+      "learning_rate": 1.6518275734645376e-05,
+      "loss": 0.3103,
+      "step": 525
+    },
+    {
+      "epoch": 2.489411764705882,
+      "grad_norm": 1.286043643951416,
+      "learning_rate": 1.628532056088669e-05,
+      "loss": 0.3348,
+      "step": 530
+    },
+    {
+      "epoch": 2.512941176470588,
+      "grad_norm": 1.1796849966049194,
+      "learning_rate": 1.6052052647106487e-05,
+      "loss": 0.3325,
+      "step": 535
+    },
+    {
+      "epoch": 2.5364705882352943,
+      "grad_norm": 1.2908035516738892,
+      "learning_rate": 1.581852875129542e-05,
+      "loss": 0.3275,
+      "step": 540
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 1.292533278465271,
+      "learning_rate": 1.5584805693728855e-05,
+      "loss": 0.3088,
+      "step": 545
+    },
+    {
+      "epoch": 2.583529411764706,
+      "grad_norm": 1.1632431745529175,
+      "learning_rate": 1.5350940343141533e-05,
+      "loss": 0.311,
+      "step": 550
+    },
+    {
+      "epoch": 2.6070588235294117,
+      "grad_norm": 1.15157151222229,
+      "learning_rate": 1.5116989602890494e-05,
+      "loss": 0.3032,
+      "step": 555
+    },
+    {
+      "epoch": 2.630588235294118,
+      "grad_norm": 1.228358507156372,
+      "learning_rate": 1.4883010397109505e-05,
+      "loss": 0.2988,
+      "step": 560
+    },
+    {
+      "epoch": 2.6541176470588237,
+      "grad_norm": 1.2099093198776245,
+      "learning_rate": 1.4649059656858472e-05,
+      "loss": 0.2979,
+      "step": 565
+    },
+    {
+      "epoch": 2.6776470588235295,
+      "grad_norm": 1.1750507354736328,
+      "learning_rate": 1.4415194306271151e-05,
+      "loss": 0.3103,
+      "step": 570
+    },
+    {
+      "epoch": 2.7011764705882353,
+      "grad_norm": 1.302628755569458,
+      "learning_rate": 1.4181471248704581e-05,
+      "loss": 0.3041,
+      "step": 575
+    },
+    {
+      "epoch": 2.724705882352941,
+      "grad_norm": 1.1675207614898682,
+      "learning_rate": 1.394794735289351e-05,
+      "loss": 0.2905,
+      "step": 580
+    },
+    {
+      "epoch": 2.748235294117647,
+      "grad_norm": 1.449485182762146,
+      "learning_rate": 1.3714679439113313e-05,
+      "loss": 0.2823,
+      "step": 585
+    },
+    {
+      "epoch": 2.7717647058823527,
+      "grad_norm": 1.1559792757034302,
+      "learning_rate": 1.3481724265354623e-05,
+      "loss": 0.2887,
+      "step": 590
+    },
+    {
+      "epoch": 2.795294117647059,
+      "grad_norm": 1.2899876832962036,
+      "learning_rate": 1.3249138513513217e-05,
+      "loss": 0.2992,
+      "step": 595
+    },
+    {
+      "epoch": 2.8188235294117647,
+      "grad_norm": 1.2164292335510254,
+      "learning_rate": 1.3016978775598299e-05,
+      "loss": 0.2889,
+      "step": 600
+    },
+    {
+      "epoch": 2.8423529411764705,
+      "grad_norm": 1.1968131065368652,
+      "learning_rate": 1.2785301539962784e-05,
+      "loss": 0.3074,
+      "step": 605
+    },
+    {
+      "epoch": 2.8658823529411763,
+      "grad_norm": 1.2663931846618652,
+      "learning_rate": 1.255416317755867e-05,
+      "loss": 0.3092,
+      "step": 610
+    },
+    {
+      "epoch": 2.8894117647058826,
+      "grad_norm": 1.3233304023742676,
+      "learning_rate": 1.23236199282211e-05,
+      "loss": 0.2966,
+      "step": 615
+    },
+    {
+      "epoch": 2.9129411764705884,
+      "grad_norm": 1.3293287754058838,
+      "learning_rate": 1.2093727886984208e-05,
+      "loss": 0.272,
+      "step": 620
+    },
+    {
+      "epoch": 2.936470588235294,
+      "grad_norm": 1.2626219987869263,
+      "learning_rate": 1.1864542990432316e-05,
+      "loss": 0.2697,
+      "step": 625
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 1.2760441303253174,
+      "learning_rate": 1.163612100308959e-05,
+      "loss": 0.284,
+      "step": 630
+    },
+    {
+      "epoch": 2.9835294117647058,
+      "grad_norm": 1.2606443166732788,
+      "learning_rate": 1.1408517503851607e-05,
+      "loss": 0.2849,
+      "step": 635
+    },
+    {
+      "epoch": 3.0047058823529413,
+      "grad_norm": 1.8812276124954224,
+      "learning_rate": 1.1181787872462075e-05,
+      "loss": 0.2844,
+      "step": 640
+    },
+    {
+      "epoch": 3.028235294117647,
+      "grad_norm": 1.25570809841156,
+      "learning_rate": 1.0955987276038016e-05,
+      "loss": 0.2286,
+      "step": 645
+    },
+    {
+      "epoch": 3.051764705882353,
+      "grad_norm": 1.2160202264785767,
+      "learning_rate": 1.0731170655646698e-05,
+      "loss": 0.2311,
+      "step": 650
+    },
+    {
+      "epoch": 3.0752941176470587,
+      "grad_norm": 1.2851207256317139,
+      "learning_rate": 1.0507392712937518e-05,
+      "loss": 0.2174,
+      "step": 655
+    },
+    {
+      "epoch": 3.0988235294117645,
+      "grad_norm": 1.2389274835586548,
+      "learning_rate": 1.0284707896832216e-05,
+      "loss": 0.2031,
+      "step": 660
+    },
+    {
+      "epoch": 3.1223529411764708,
+      "grad_norm": 1.2528103590011597,
+      "learning_rate": 1.0063170390276499e-05,
+      "loss": 0.1952,
+      "step": 665
+    },
+    {
+      "epoch": 3.1458823529411766,
+      "grad_norm": 1.441513180732727,
+      "learning_rate": 9.842834097056464e-06,
+      "loss": 0.2002,
+      "step": 670
+    },
+    {
+      "epoch": 3.1694117647058824,
+      "grad_norm": 1.1801048517227173,
+      "learning_rate": 9.62375262868286e-06,
+      "loss": 0.2107,
+      "step": 675
+    },
+    {
+      "epoch": 3.192941176470588,
+      "grad_norm": 1.1453895568847656,
+      "learning_rate": 9.405979291346556e-06,
+      "loss": 0.2181,
+      "step": 680
+    },
+    {
+      "epoch": 3.216470588235294,
+      "grad_norm": 1.1607674360275269,
+      "learning_rate": 9.18956707294821e-06,
+      "loss": 0.2023,
+      "step": 685
+    },
+    {
+      "epoch": 3.24,
+      "grad_norm": 1.2503108978271484,
+      "learning_rate": 8.974568630205462e-06,
+      "loss": 0.2022,
+      "step": 690
+    },
+    {
+      "epoch": 3.263529411764706,
+      "grad_norm": 1.3469291925430298,
+      "learning_rate": 8.76103627584063e-06,
+      "loss": 0.2266,
+      "step": 695
+    },
+    {
+      "epoch": 3.287058823529412,
+      "grad_norm": 1.2448289394378662,
+      "learning_rate": 8.549021965852198e-06,
+      "loss": 0.2261,
+      "step": 700
+    },
+    {
+      "epoch": 3.3105882352941176,
+      "grad_norm": 1.33231520652771,
+      "learning_rate": 8.338577286873002e-06,
+      "loss": 0.1907,
+      "step": 705
+    },
+    {
+      "epoch": 3.3341176470588234,
+      "grad_norm": 1.1715577840805054,
+      "learning_rate": 8.129753443618391e-06,
+      "loss": 0.2045,
+      "step": 710
+    },
+    {
+      "epoch": 3.357647058823529,
+      "grad_norm": 1.2334033250808716,
+      "learning_rate": 7.922601246427203e-06,
+      "loss": 0.1909,
+      "step": 715
+    },
+    {
+      "epoch": 3.3811764705882354,
+      "grad_norm": 1.5290734767913818,
+      "learning_rate": 7.717171098898803e-06,
+      "loss": 0.204,
+      "step": 720
+    },
+    {
+      "epoch": 3.4047058823529412,
+      "grad_norm": 1.4729763269424438,
+      "learning_rate": 7.513512985629037e-06,
+      "loss": 0.209,
+      "step": 725
+    },
+    {
+      "epoch": 3.428235294117647,
+      "grad_norm": 1.1317057609558105,
+      "learning_rate": 7.311676460048128e-06,
+      "loss": 0.1913,
+      "step": 730
+    },
+    {
+      "epoch": 3.451764705882353,
+      "grad_norm": 1.2160744667053223,
+      "learning_rate": 7.1117106323634886e-06,
+      "loss": 0.1897,
+      "step": 735
+    },
+    {
+      "epoch": 3.4752941176470586,
+      "grad_norm": 1.176538348197937,
+      "learning_rate": 6.913664157610425e-06,
+      "loss": 0.2003,
+      "step": 740
+    },
+    {
+      "epoch": 3.498823529411765,
+      "grad_norm": 1.3141257762908936,
+      "learning_rate": 6.717585223813497e-06,
+      "loss": 0.1923,
+      "step": 745
+    },
+    {
+      "epoch": 3.5223529411764707,
+      "grad_norm": 1.2836600542068481,
+      "learning_rate": 6.523521540261626e-06,
+      "loss": 0.1991,
+      "step": 750
+    },
+    {
+      "epoch": 3.5458823529411765,
+      "grad_norm": 1.242904543876648,
+      "learning_rate": 6.331520325899561e-06,
+      "loss": 0.2008,
+      "step": 755
+    },
+    {
+      "epoch": 3.5694117647058823,
+      "grad_norm": 1.1963882446289062,
+      "learning_rate": 6.14162829783878e-06,
+      "loss": 0.1804,
+      "step": 760
+    },
+    {
+      "epoch": 3.592941176470588,
+      "grad_norm": 1.0519522428512573,
+      "learning_rate": 5.953891659990377e-06,
+      "loss": 0.1851,
+      "step": 765
+    },
+    {
+      "epoch": 3.616470588235294,
+      "grad_norm": 1.3625632524490356,
+      "learning_rate": 5.768356091822926e-06,
+      "loss": 0.1831,
+      "step": 770
+    },
+    {
+      "epoch": 3.64,
+      "grad_norm": 1.5194118022918701,
+      "learning_rate": 5.585066737247876e-06,
+      "loss": 0.1943,
+      "step": 775
+    },
+    {
+      "epoch": 3.663529411764706,
+      "grad_norm": 1.3785059452056885,
+      "learning_rate": 5.404068193635283e-06,
+      "loss": 0.1795,
+      "step": 780
+    },
+    {
+      "epoch": 3.6870588235294117,
+      "grad_norm": 1.4833396673202515,
+      "learning_rate": 5.225404500962516e-06,
+      "loss": 0.1856,
+      "step": 785
+    },
+    {
+      "epoch": 3.7105882352941175,
+      "grad_norm": 1.3318918943405151,
+      "learning_rate": 5.04911913109862e-06,
+      "loss": 0.1741,
+      "step": 790
+    },
+    {
+      "epoch": 3.7341176470588238,
+      "grad_norm": 1.3155924081802368,
+      "learning_rate": 4.875254977226845e-06,
+      "loss": 0.1782,
+      "step": 795
+    },
+    {
+      "epoch": 3.7576470588235296,
+      "grad_norm": 2.2067034244537354,
+      "learning_rate": 4.703854343408042e-06,
+      "loss": 0.1906,
+      "step": 800
+    },
+    {
+      "epoch": 3.7811764705882354,
+      "grad_norm": 1.2327618598937988,
+      "learning_rate": 4.534958934287378e-06,
+      "loss": 0.1911,
+      "step": 805
+    },
+    {
+      "epoch": 3.804705882352941,
+      "grad_norm": 1.2742501497268677,
+      "learning_rate": 4.368609844946873e-06,
+      "loss": 0.1613,
+      "step": 810
+    },
+    {
+      "epoch": 3.828235294117647,
+      "grad_norm": 1.2024046182632446,
+      "learning_rate": 4.2048475509063065e-06,
+      "loss": 0.1459,
+      "step": 815
+    },
+    {
+      "epoch": 3.8517647058823528,
+      "grad_norm": 1.3519951105117798,
+      "learning_rate": 4.043711898274828e-06,
+      "loss": 0.1639,
+      "step": 820
+    },
+    {
+      "epoch": 3.8752941176470586,
+      "grad_norm": 1.2287205457687378,
+      "learning_rate": 3.885242094055737e-06,
+      "loss": 0.1984,
+      "step": 825
+    },
+    {
+      "epoch": 3.898823529411765,
+      "grad_norm": 1.3539725542068481,
+      "learning_rate": 3.729476696606759e-06,
+      "loss": 0.1934,
+      "step": 830
+    },
+    {
+      "epoch": 3.9223529411764706,
+      "grad_norm": 1.4263674020767212,
+      "learning_rate": 3.5764536062581575e-06,
+      "loss": 0.1918,
+      "step": 835
+    },
+    {
+      "epoch": 3.9458823529411764,
+      "grad_norm": 1.1720516681671143,
+      "learning_rate": 3.4262100560909387e-06,
+      "loss": 0.1756,
+      "step": 840
+    },
+    {
+      "epoch": 3.969411764705882,
+      "grad_norm": 1.2408292293548584,
+      "learning_rate": 3.2787826028774364e-06,
+      "loss": 0.1622,
+      "step": 845
+    },
+    {
+      "epoch": 3.9929411764705884,
+      "grad_norm": 1.4117648601531982,
+      "learning_rate": 3.1342071181864114e-06,
+      "loss": 0.1714,
+      "step": 850
+    },
+    {
+      "epoch": 4.014117647058823,
+      "grad_norm": 1.0455414056777954,
+      "learning_rate": 2.9925187796549265e-06,
+      "loss": 0.1591,
+      "step": 855
+    },
+    {
+      "epoch": 4.03764705882353,
+      "grad_norm": 1.2188994884490967,
+      "learning_rate": 2.8537520624290146e-06,
+      "loss": 0.1361,
+      "step": 860
+    },
+    {
+      "epoch": 4.061176470588236,
+      "grad_norm": 1.3623497486114502,
+      "learning_rate": 2.7179407307753155e-06,
+      "loss": 0.1447,
+      "step": 865
+    },
+    {
+      "epoch": 4.084705882352941,
+      "grad_norm": 1.0604227781295776,
+      "learning_rate": 2.585117829865648e-06,
+      "loss": 0.1715,
+      "step": 870
+    },
+    {
+      "epoch": 4.108235294117647,
+      "grad_norm": 1.1241997480392456,
+      "learning_rate": 2.4553156777365837e-06,
+      "loss": 0.1493,
+      "step": 875
+    },
+    {
+      "epoch": 4.131764705882353,
+      "grad_norm": 1.1583119630813599,
+      "learning_rate": 2.3285658574259146e-06,
+      "loss": 0.1488,
+      "step": 880
+    },
+    {
+      "epoch": 4.155294117647059,
+      "grad_norm": 1.309241533279419,
+      "learning_rate": 2.204899209287977e-06,
+      "loss": 0.1384,
+      "step": 885
+    },
+    {
+      "epoch": 4.178823529411765,
+      "grad_norm": 1.1592360734939575,
+      "learning_rate": 2.0843458234896668e-06,
+      "loss": 0.1368,
+      "step": 890
+    },
+    {
+      "epoch": 4.20235294117647,
+      "grad_norm": 1.1202175617218018,
+      "learning_rate": 1.9669350326890175e-06,
+      "loss": 0.1508,
+      "step": 895
+    },
+    {
+      "epoch": 4.225882352941176,
+      "grad_norm": 1.2421292066574097,
+      "learning_rate": 1.8526954048980499e-06,
+      "loss": 0.1394,
+      "step": 900
+    },
+    {
+      "epoch": 4.249411764705882,
+      "grad_norm": 1.3553770780563354,
+      "learning_rate": 1.7416547365317176e-06,
+      "loss": 0.1273,
+      "step": 905
+    },
+    {
+      "epoch": 4.272941176470589,
+      "grad_norm": 1.1461013555526733,
+      "learning_rate": 1.6338400456445557e-06,
+      "loss": 0.1524,
+      "step": 910
+    },
+    {
+      "epoch": 4.2964705882352945,
+      "grad_norm": 1.1495455503463745,
+      "learning_rate": 1.529277565356752e-06,
+      "loss": 0.1371,
+      "step": 915
+    },
+    {
+      "epoch": 4.32,
+      "grad_norm": 1.2553972005844116,
+      "learning_rate": 1.4279927374711692e-06,
+      "loss": 0.1437,
+      "step": 920
+    },
+    {
+      "epoch": 4.343529411764706,
+      "grad_norm": 1.0836520195007324,
+      "learning_rate": 1.3300102062829521e-06,
+      "loss": 0.1345,
+      "step": 925
+    },
+    {
+      "epoch": 4.367058823529412,
+      "grad_norm": 1.1315704584121704,
+      "learning_rate": 1.235353812583142e-06,
+      "loss": 0.1238,
+      "step": 930
+    },
+    {
+      "epoch": 4.390588235294118,
+      "grad_norm": 1.0683834552764893,
+      "learning_rate": 1.1440465878578156e-06,
+      "loss": 0.1242,
+      "step": 935
+    },
+    {
+      "epoch": 4.4141176470588235,
+      "grad_norm": 1.164872169494629,
+      "learning_rate": 1.056110748684137e-06,
+      "loss": 0.1299,
+      "step": 940
+    },
+    {
+      "epoch": 4.437647058823529,
+      "grad_norm": 1.1151331663131714,
+      "learning_rate": 9.715676913246984e-07,
+      "loss": 0.1434,
+      "step": 945
+    },
+    {
+      "epoch": 4.461176470588235,
+      "grad_norm": 1.319403052330017,
+      "learning_rate": 8.904379865214479e-07,
+      "loss": 0.1291,
+      "step": 950
+    },
+    {
+      "epoch": 4.484705882352941,
+      "grad_norm": 1.151589274406433,
+      "learning_rate": 8.127413744904805e-07,
+      "loss": 0.1323,
+      "step": 955
+    },
+    {
+      "epoch": 4.508235294117647,
+      "grad_norm": 1.234084129333496,
+      "learning_rate": 7.384967601189203e-07,
+      "loss": 0.1319,
+      "step": 960
+    },
+    {
+      "epoch": 4.5317647058823525,
+      "grad_norm": 1.1618750095367432,
+      "learning_rate": 6.677222083650347e-07,
+      "loss": 0.1353,
+      "step": 965
+    },
+    {
+      "epoch": 4.555294117647059,
+      "grad_norm": 1.0726864337921143,
+      "learning_rate": 6.004349398627362e-07,
+      "loss": 0.1174,
+      "step": 970
+    },
+    {
+      "epoch": 4.578823529411765,
+      "grad_norm": 1.2831597328186035,
+      "learning_rate": 5.366513267315082e-07,
+      "loss": 0.1299,
+      "step": 975
+    },
+    {
+      "epoch": 4.602352941176471,
+      "grad_norm": 1.2213561534881592,
+      "learning_rate": 4.7638688859280186e-07,
+      "loss": 0.146,
+      "step": 980
+    },
+    {
+      "epoch": 4.625882352941177,
+      "grad_norm": 1.2630505561828613,
+      "learning_rate": 4.1965628879383877e-07,
+      "loss": 0.1449,
+      "step": 985
+    },
+    {
+      "epoch": 4.649411764705882,
+      "grad_norm": 1.1484320163726807,
+      "learning_rate": 3.664733308397861e-07,
+      "loss": 0.1467,
+      "step": 990
+    },
+    {
+      "epoch": 4.672941176470588,
+      "grad_norm": 1.234032392501831,
+      "learning_rate": 3.168509550351234e-07,
+      "loss": 0.1184,
+      "step": 995
+    },
+    {
+      "epoch": 4.696470588235294,
+      "grad_norm": 1.1044232845306396,
+      "learning_rate": 2.7080123533505807e-07,
+      "loss": 0.1214,
+      "step": 1000
+    },
+    {
+      "epoch": 4.72,
+      "grad_norm": 1.1763945817947388,
+      "learning_rate": 2.2833537640772339e-07,
+      "loss": 0.1293,
+      "step": 1005
+    },
+    {
+      "epoch": 4.7435294117647056,
+      "grad_norm": 1.1263889074325562,
+      "learning_rate": 1.894637109078978e-07,
+      "loss": 0.1245,
+      "step": 1010
+    },
+    {
+      "epoch": 4.767058823529412,
+      "grad_norm": 1.1052639484405518,
+      "learning_rate": 1.541956969628927e-07,
+      "loss": 0.1499,
+      "step": 1015
+    },
+    {
+      "epoch": 4.790588235294118,
+      "grad_norm": 1.1130540370941162,
+      "learning_rate": 1.2253991587124313e-07,
+      "loss": 0.1284,
+      "step": 1020
+    },
+    {
+      "epoch": 4.814117647058824,
+      "grad_norm": 1.2002348899841309,
+      "learning_rate": 9.450407001471651e-08,
+      "loss": 0.1216,
+      "step": 1025
+    },
+    {
+      "epoch": 4.83764705882353,
+      "grad_norm": 1.226212739944458,
+      "learning_rate": 7.009498098421163e-08,
+      "loss": 0.1402,
+      "step": 1030
+    },
+    {
+      "epoch": 4.861176470588235,
+      "grad_norm": 1.1810014247894287,
+      "learning_rate": 4.931858791993815e-08,
+      "loss": 0.1288,
+      "step": 1035
+    },
+    {
+      "epoch": 4.884705882352941,
+      "grad_norm": 1.086016297340393,
+      "learning_rate": 3.2179946066335786e-08,
+      "loss": 0.1251,
+      "step": 1040
+    },
+    {
+      "epoch": 4.908235294117647,
+      "grad_norm": 1.1115578413009644,
+      "learning_rate": 1.8683225542034943e-08,
+      "loss": 0.1162,
+      "step": 1045
+    },
+    {
+      "epoch": 4.931764705882353,
+      "grad_norm": 1.2402100563049316,
+      "learning_rate": 8.831710325211684e-09,
+      "loss": 0.1363,
+      "step": 1050
+    },
+    {
+      "epoch": 4.955294117647059,
+      "grad_norm": 1.0764867067337036,
+      "learning_rate": 2.627797454528569e-09,
+      "loss": 0.1133,
+      "step": 1055
+    },
+    {
+      "epoch": 4.978823529411764,
+      "grad_norm": 1.103387475013733,
+      "learning_rate": 7.299644590286203e-11,
+      "loss": 0.1282,
+      "step": 1060
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1060,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.3555482874376356e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec816804797af459a9efb4ac633039ac077e4e68b7e00381de926e936f3af410
+size 7800

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-1060/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/base/4_128_e5_3e-5/checkpoint-213/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/base/4_128_e5_3e-5/checkpoint-213/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",
+    "up_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91366994454c646c3708d8a0544b040f8d4fa6e0fced8067f3978cdcf431cd2b
+size 791751704

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/latest

@@ -0,0 +1 @@
+global_step212

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1bf35940968a3391f2d0def50ed247eb58d6304303074cc24d73590c4416a2f
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba153d49befc7bd8ec55d8cfca64faedd68c870ff53391de4b164eb4bbf63ff0
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9c52d5c3ca297742c4a045546cdce5d839cfb4476a1a4f12c120ca695cfb264
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67660c7d06c2c368372ee5dc1cf00ff5643dedce669582e913e8836067c4a1e9
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:988843a7b0ee00d328c37508ad3aeb334aba6bd955c1401c6c3ace48cc5a9359
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:728a790dad9d96b44a0b34b3873ed83259646eaa62be26724a75e791a09a70c6
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc6de7d1451c795b35e20a9f03d160d45c3ae3269cef78b3fbbe0febfb0a0580
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5272f3264a14cb5210e9db6c0ab492588d4ff759c57978426fabc67a8b7d39
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a66c53e5c268d61fb9886f0f822ef8ddae37edb073941126ada8fb354756bd3
+size 1064

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/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/base/4_128_e5_3e-5/checkpoint-213/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/base/4_128_e5_3e-5/checkpoint-213/trainer_state.json

@@ -0,0 +1,328 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 213,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.023529411764705882,
+      "grad_norm": 0.8258791565895081,
+      "learning_rate": 2.2641509433962262e-06,
+      "loss": 1.3394,
+      "step": 5
+    },
+    {
+      "epoch": 0.047058823529411764,
+      "grad_norm": 0.5909937024116516,
+      "learning_rate": 5.094339622641509e-06,
+      "loss": 1.3138,
+      "step": 10
+    },
+    {
+      "epoch": 0.07058823529411765,
+      "grad_norm": 0.5355730652809143,
+      "learning_rate": 7.924528301886793e-06,
+      "loss": 1.3315,
+      "step": 15
+    },
+    {
+      "epoch": 0.09411764705882353,
+      "grad_norm": 0.5410304665565491,
+      "learning_rate": 1.0754716981132076e-05,
+      "loss": 1.3214,
+      "step": 20
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 0.4087582230567932,
+      "learning_rate": 1.358490566037736e-05,
+      "loss": 1.2642,
+      "step": 25
+    },
+    {
+      "epoch": 0.1411764705882353,
+      "grad_norm": 0.3983219861984253,
+      "learning_rate": 1.6415094339622643e-05,
+      "loss": 1.2698,
+      "step": 30
+    },
+    {
+      "epoch": 0.16470588235294117,
+      "grad_norm": 0.39306771755218506,
+      "learning_rate": 1.9245283018867924e-05,
+      "loss": 1.2605,
+      "step": 35
+    },
+    {
+      "epoch": 0.18823529411764706,
+      "grad_norm": 0.3943307101726532,
+      "learning_rate": 2.2075471698113208e-05,
+      "loss": 1.2468,
+      "step": 40
+    },
+    {
+      "epoch": 0.21176470588235294,
+      "grad_norm": 0.34280142188072205,
+      "learning_rate": 2.4905660377358492e-05,
+      "loss": 1.2487,
+      "step": 45
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 0.39148157835006714,
+      "learning_rate": 2.7735849056603773e-05,
+      "loss": 1.2446,
+      "step": 50
+    },
+    {
+      "epoch": 0.25882352941176473,
+      "grad_norm": 0.44964322447776794,
+      "learning_rate": 2.99999270035541e-05,
+      "loss": 1.2281,
+      "step": 55
+    },
+    {
+      "epoch": 0.2823529411764706,
+      "grad_norm": 0.40044668316841125,
+      "learning_rate": 2.999737220254547e-05,
+      "loss": 1.1738,
+      "step": 60
+    },
+    {
+      "epoch": 0.3058823529411765,
+      "grad_norm": 0.45404553413391113,
+      "learning_rate": 2.999116828967479e-05,
+      "loss": 1.1659,
+      "step": 65
+    },
+    {
+      "epoch": 0.32941176470588235,
+      "grad_norm": 0.4405958652496338,
+      "learning_rate": 2.9981316774457966e-05,
+      "loss": 1.1103,
+      "step": 70
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 0.43754836916923523,
+      "learning_rate": 2.9967820053933663e-05,
+      "loss": 1.1569,
+      "step": 75
+    },
+    {
+      "epoch": 0.3764705882352941,
+      "grad_norm": 0.48351433873176575,
+      "learning_rate": 2.995068141208006e-05,
+      "loss": 1.0878,
+      "step": 80
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.45371121168136597,
+      "learning_rate": 2.992990501901579e-05,
+      "loss": 1.0451,
+      "step": 85
+    },
+    {
+      "epoch": 0.4235294117647059,
+      "grad_norm": 0.4767340123653412,
+      "learning_rate": 2.9905495929985282e-05,
+      "loss": 1.0701,
+      "step": 90
+    },
+    {
+      "epoch": 0.4470588235294118,
+      "grad_norm": 0.6476609110832214,
+      "learning_rate": 2.987746008412876e-05,
+      "loss": 1.0156,
+      "step": 95
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 0.4992612302303314,
+      "learning_rate": 2.984580430303711e-05,
+      "loss": 1.0879,
+      "step": 100
+    },
+    {
+      "epoch": 0.49411764705882355,
+      "grad_norm": 0.5524128079414368,
+      "learning_rate": 2.98105362890921e-05,
+      "loss": 1.0219,
+      "step": 105
+    },
+    {
+      "epoch": 0.5176470588235295,
+      "grad_norm": 0.7320187091827393,
+      "learning_rate": 2.9771664623592276e-05,
+      "loss": 1.0294,
+      "step": 110
+    },
+    {
+      "epoch": 0.5411764705882353,
+      "grad_norm": 0.5769292116165161,
+      "learning_rate": 2.9729198764664944e-05,
+      "loss": 0.9888,
+      "step": 115
+    },
+    {
+      "epoch": 0.5647058823529412,
+      "grad_norm": 0.6180774569511414,
+      "learning_rate": 2.968314904496488e-05,
+      "loss": 1.0136,
+      "step": 120
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 0.6959433555603027,
+      "learning_rate": 2.9633526669160217e-05,
+      "loss": 1.0116,
+      "step": 125
+    },
+    {
+      "epoch": 0.611764705882353,
+      "grad_norm": 0.7187851071357727,
+      "learning_rate": 2.9580343711206162e-05,
+      "loss": 0.9455,
+      "step": 130
+    },
+    {
+      "epoch": 0.6352941176470588,
+      "grad_norm": 0.7223728895187378,
+      "learning_rate": 2.9523613111407196e-05,
+      "loss": 0.9332,
+      "step": 135
+    },
+    {
+      "epoch": 0.6588235294117647,
+      "grad_norm": 0.6712695956230164,
+      "learning_rate": 2.946334867326849e-05,
+      "loss": 0.9047,
+      "step": 140
+    },
+    {
+      "epoch": 0.6823529411764706,
+      "grad_norm": 0.6632208824157715,
+      "learning_rate": 2.939956506013727e-05,
+      "loss": 0.9201,
+      "step": 145
+    },
+    {
+      "epoch": 0.7058823529411765,
+      "grad_norm": 0.7727546691894531,
+      "learning_rate": 2.933227779163497e-05,
+      "loss": 0.8752,
+      "step": 150
+    },
+    {
+      "epoch": 0.7294117647058823,
+      "grad_norm": 0.7360590100288391,
+      "learning_rate": 2.926150323988108e-05,
+      "loss": 0.8369,
+      "step": 155
+    },
+    {
+      "epoch": 0.7529411764705882,
+      "grad_norm": 0.7336561679840088,
+      "learning_rate": 2.9187258625509518e-05,
+      "loss": 0.8651,
+      "step": 160
+    },
+    {
+      "epoch": 0.7764705882352941,
+      "grad_norm": 0.7670441269874573,
+      "learning_rate": 2.9109562013478553e-05,
+      "loss": 0.8366,
+      "step": 165
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.7336223721504211,
+      "learning_rate": 2.9028432308675305e-05,
+      "loss": 0.9056,
+      "step": 170
+    },
+    {
+      "epoch": 0.8235294117647058,
+      "grad_norm": 0.8804784417152405,
+      "learning_rate": 2.8943889251315867e-05,
+      "loss": 0.8196,
+      "step": 175
+    },
+    {
+      "epoch": 0.8470588235294118,
+      "grad_norm": 0.8821813464164734,
+      "learning_rate": 2.8855953412142184e-05,
+      "loss": 0.7661,
+      "step": 180
+    },
+    {
+      "epoch": 0.8705882352941177,
+      "grad_norm": 1.065437912940979,
+      "learning_rate": 2.876464618741686e-05,
+      "loss": 0.8047,
+      "step": 185
+    },
+    {
+      "epoch": 0.8941176470588236,
+      "grad_norm": 0.957807183265686,
+      "learning_rate": 2.866998979371705e-05,
+      "loss": 0.8138,
+      "step": 190
+    },
+    {
+      "epoch": 0.9176470588235294,
+      "grad_norm": 0.8701325058937073,
+      "learning_rate": 2.857200726252883e-05,
+      "loss": 0.7415,
+      "step": 195
+    },
+    {
+      "epoch": 0.9411764705882353,
+      "grad_norm": 0.8715978860855103,
+      "learning_rate": 2.847072243464325e-05,
+      "loss": 0.8062,
+      "step": 200
+    },
+    {
+      "epoch": 0.9647058823529412,
+      "grad_norm": 0.8279728889465332,
+      "learning_rate": 2.8366159954355444e-05,
+      "loss": 0.7496,
+      "step": 205
+    },
+    {
+      "epoch": 0.9882352941176471,
+      "grad_norm": 0.8706197142601013,
+      "learning_rate": 2.8258345263468284e-05,
+      "loss": 0.721,
+      "step": 210
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1060,
+  "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": 4.739250739006669e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec816804797af459a9efb4ac633039ac077e4e68b7e00381de926e936f3af410
+size 7800

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-213/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/base/4_128_e5_3e-5/checkpoint-426/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/base/4_128_e5_3e-5/checkpoint-426/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",
+    "up_proj",
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-426/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:038874b69c5382adbe261944f3df5f070113304a7d177338cb8963a4e4dd7aa2
+size 791751704

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-426/latest

@@ -0,0 +1 @@
+global_step425

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-426/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31a2274bdef4569895ac0f47d7534bce49d92092f1575968a821bd0160333be5
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-426/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e362d905f65f04384f9ba2bd12fb66747e193f9b5966237ead450d4ed5a854b3
+size 15920

hotpotqa/not_grounded/base/4_128_e5_3e-5/checkpoint-426/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13ec2de484c70f85aa875e1a82504908f28966df22dfa7634246c880d4d14a91
+size 15920