Instructions to use adpretko/train-riscv-O2_epoch1and2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use adpretko/train-riscv-O2_epoch1and2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="adpretko/train-riscv-O2_epoch1and2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("adpretko/train-riscv-O2_epoch1and2")
model = AutoModelForCausalLM.from_pretrained("adpretko/train-riscv-O2_epoch1and2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use adpretko/train-riscv-O2_epoch1and2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "adpretko/train-riscv-O2_epoch1and2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adpretko/train-riscv-O2_epoch1and2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/adpretko/train-riscv-O2_epoch1and2

SGLang

How to use adpretko/train-riscv-O2_epoch1and2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "adpretko/train-riscv-O2_epoch1and2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adpretko/train-riscv-O2_epoch1and2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "adpretko/train-riscv-O2_epoch1and2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adpretko/train-riscv-O2_epoch1and2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use adpretko/train-riscv-O2_epoch1and2 with Docker Model Runner:
```
docker model run hf.co/adpretko/train-riscv-O2_epoch1and2
```

adpretko commited on Nov 3, 2025

Commit

fb67044

verified ·

1 Parent(s): bc0faa7

Upload fine-tuned riscv-O2 model

Browse files

Files changed (37) hide show

.gitattributes +1 -0
README.md +59 -0
checkpoint-3800/added_tokens.json +24 -0
checkpoint-3800/chat_template.jinja +54 -0
checkpoint-3800/config.json +58 -0
checkpoint-3800/generation_config.json +14 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt +3 -0
checkpoint-3800/global_step3799/mp_rank_00_model_states.pt +3 -0
checkpoint-3800/latest +1 -0
checkpoint-3800/merges.txt +0 -0
checkpoint-3800/model.safetensors +3 -0
checkpoint-3800/rng_state_0.pth +3 -0
checkpoint-3800/rng_state_1.pth +3 -0
checkpoint-3800/rng_state_2.pth +3 -0
checkpoint-3800/rng_state_3.pth +3 -0
checkpoint-3800/rng_state_4.pth +3 -0
checkpoint-3800/rng_state_5.pth +3 -0
checkpoint-3800/rng_state_6.pth +3 -0
checkpoint-3800/rng_state_7.pth +3 -0
checkpoint-3800/scheduler.pt +3 -0
checkpoint-3800/special_tokens_map.json +31 -0
checkpoint-3800/tokenizer.json +3 -0
checkpoint-3800/tokenizer_config.json +208 -0
checkpoint-3800/trainer_state.json +2694 -0
checkpoint-3800/training_args.bin +3 -0
checkpoint-3800/vocab.json +0 -0
checkpoint-3800/zero_to_fp32.py +760 -0
generation_config.json +14 -0
model.safetensors +1 -1
trainer_log.jsonl +9 -0

.gitattributes CHANGED Viewed

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text

 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-3800/tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md ADDED Viewed

	@@ -0,0 +1,59 @@

+---
+library_name: transformers
+base_model: Qwen/Qwen2.5-1.5B-Instruct
+tags:
+- llama-factory
+- generated_from_trainer
+model-index:
+- name: train-riscv-O2_epoch1and2
+  results: []
+---
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+# train-riscv-O2_epoch1and2
+This model is a fine-tuned version of [saves/train-riscv-O2_epoch1and2/checkpoint-2800](https://huggingface.co/saves/train-riscv-O2_epoch1and2/checkpoint-2800) on the None dataset.
+## Model description
+More information needed
+## Intended uses & limitations
+More information needed
+## Training and evaluation data
+More information needed
+## Training procedure
+### Training hyperparameters
+The following hyperparameters were used during training:
+- learning_rate: 2e-05
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 8
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 512
+- total_eval_batch_size: 64
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_ratio: 0.1
+- num_epochs: 2.0
+### Training results
+### Framework versions
+- Transformers 4.55.0
+- Pytorch 2.8.0+rocm6.3
+- Datasets 3.6.0
+- Tokenizers 0.21.1

checkpoint-3800/added_tokens.json ADDED Viewed

	@@ -0,0 +1,24 @@

+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-3800/chat_template.jinja ADDED Viewed

	@@ -0,0 +1,54 @@

+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

checkpoint-3800/config.json ADDED Viewed

	@@ -0,0 +1,58 @@

+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.55.0",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

checkpoint-3800/generation_config.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.55.0"
+}

checkpoint-3800/global_step3799/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:73aa60ee27ca8fdaa7c035ac068fe58235cabccc7b127e759a7c664e6a782dde
+size 2315578373

checkpoint-3800/global_step3799/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7fa0dbc2d9a725a5645aa751f4a79de0f9df5d0d6274002340fb8bc0043352c2
+size 2315580229

checkpoint-3800/global_step3799/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:3ab0fa72eb491d80b3a0dc3159a24909a4ae7c908260f683643794e90f1f60af
+size 2315580485

checkpoint-3800/global_step3799/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:a0e3e6646a2a54768db511a4873904ebbe2abf5f607aa31b0ff479436e2be6ab
+size 2315580485

checkpoint-3800/global_step3799/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:c9178dd494a4512e2fd1840b28f33a835485802c723efc54c63488010d337dee
+size 2315580485

checkpoint-3800/global_step3799/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:4aaa5f086465e9c2d514bad0240dc1660045970c8b19fa087bea0d80b581957d
+size 2315580485

checkpoint-3800/global_step3799/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:628a7b341047bf3052dd6ca65aa979fcbbab4b84807a1a29986947e3e8a3b0fc
+size 2315580613

checkpoint-3800/global_step3799/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7bbcb8d1e4c44b92046e4d83ecd26dcc86086f407b012d4149b21a9274b3ad81
+size 2315580357

checkpoint-3800/global_step3799/mp_rank_00_model_states.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6fd4b56bb2d10b8cbabf31d7dcdace8c920ac000a0a8f7b82484d25621d70bbe
+size 3087519461

checkpoint-3800/latest ADDED Viewed

	@@ -0,0 +1 @@


1	+ global_step3799

checkpoint-3800/merges.txt ADDED Viewed

The diff for this file is too large to render. See raw diff

checkpoint-3800/model.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1b5efb4b71ab5f052dbf8615c5024a2e44383c8a3c81433fbbf5f0a65110a795
+size 3554214752

checkpoint-3800/rng_state_0.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:b0418786c564bb001fc5f4215375b1a99ed3ea3f30771057e1d8682214a0108f
+size 16389

checkpoint-3800/rng_state_1.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:35deaabe10396919e53b2274736981b26bc690165c8839845f178b355a6a7588
+size 16389

checkpoint-3800/rng_state_2.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1c5deee061a76a9e05b4634f5abad786b4042e744cc4c156a2ab76dc3c94fcac
+size 16389

checkpoint-3800/rng_state_3.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:13579f7ca365ea6f9cef83929c34833cd97f9b5764a434989b4e3e2db87dce56
+size 16389

checkpoint-3800/rng_state_4.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:073d0b1d7e2e3930e5c29fc418e6246462ab2a00396c3a86e1448d373503f1b2
+size 16389

checkpoint-3800/rng_state_5.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7c60731015889939e665a308a0933ac6c327734a4699cf5615217db0cfaf03ca
+size 16389

checkpoint-3800/rng_state_6.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:416648e7a35618f28d73625751a8fc2ad066533414d19f6fc0c39932429023fc
+size 16389

checkpoint-3800/rng_state_7.pth ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:619b5c49446c94dca2b233ad17098e42f7d58a945c2eced1aa23e50469a75431
+size 16389

checkpoint-3800/scheduler.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:340752ce40761cf7bbd35a29594a7a48c8e621c1da71a043752000d166648fa0
+size 1465

checkpoint-3800/special_tokens_map.json ADDED Viewed

	@@ -0,0 +1,31 @@

+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-3800/tokenizer.json ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoint-3800/tokenizer_config.json ADDED Viewed

	@@ -0,0 +1,208 @@

+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-3800/trainer_state.json ADDED Viewed

	@@ -0,0 +1,2694 @@

+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9558615364817913,
+  "eval_steps": 500,
+  "global_step": 3800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0051473426843392095,
+      "grad_norm": 5.310028553009033,
+      "learning_rate": 4.6272493573264783e-07,
+      "loss": 0.5767,
+      "step": 10
+    },
+    {
+      "epoch": 0.010294685368678419,
+      "grad_norm": 3.2451159954071045,
+      "learning_rate": 9.768637532133676e-07,
+      "loss": 0.5436,
+      "step": 20
+    },
+    {
+      "epoch": 0.01544202805301763,
+      "grad_norm": 2.6238412857055664,
+      "learning_rate": 1.4910025706940876e-06,
+      "loss": 0.4248,
+      "step": 30
+    },
+    {
+      "epoch": 0.020589370737356838,
+      "grad_norm": 1.1435836553573608,
+      "learning_rate": 2.0051413881748076e-06,
+      "loss": 0.2565,
+      "step": 40
+    },
+    {
+      "epoch": 0.02573671342169605,
+      "grad_norm": 0.6369749307632446,
+      "learning_rate": 2.519280205655527e-06,
+      "loss": 0.1546,
+      "step": 50
+    },
+    {
+      "epoch": 0.03088405610603526,
+      "grad_norm": 0.31389811635017395,
+      "learning_rate": 3.033419023136247e-06,
+      "loss": 0.1188,
+      "step": 60
+    },
+    {
+      "epoch": 0.036031398790374466,
+      "grad_norm": 0.23692786693572998,
+      "learning_rate": 3.547557840616967e-06,
+      "loss": 0.1005,
+      "step": 70
+    },
+    {
+      "epoch": 0.041178741474713676,
+      "grad_norm": 0.17490938305854797,
+      "learning_rate": 4.0616966580976866e-06,
+      "loss": 0.0885,
+      "step": 80
+    },
+    {
+      "epoch": 0.04632608415905289,
+      "grad_norm": 0.16715744137763977,
+      "learning_rate": 4.575835475578407e-06,
+      "loss": 0.078,
+      "step": 90
+    },
+    {
+      "epoch": 0.0514734268433921,
+      "grad_norm": 0.168691024184227,
+      "learning_rate": 5.0899742930591265e-06,
+      "loss": 0.0728,
+      "step": 100
+    },
+    {
+      "epoch": 0.05662076952773131,
+      "grad_norm": 0.1685110628604889,
+      "learning_rate": 5.604113110539846e-06,
+      "loss": 0.0652,
+      "step": 110
+    },
+    {
+      "epoch": 0.06176811221207052,
+      "grad_norm": 0.14894704520702362,
+      "learning_rate": 6.1182519280205664e-06,
+      "loss": 0.0611,
+      "step": 120
+    },
+    {
+      "epoch": 0.06691545489640972,
+      "grad_norm": 0.17869767546653748,
+      "learning_rate": 6.632390745501286e-06,
+      "loss": 0.0578,
+      "step": 130
+    },
+    {
+      "epoch": 0.07206279758074893,
+      "grad_norm": 0.2117505520582199,
+      "learning_rate": 7.1465295629820055e-06,
+      "loss": 0.0565,
+      "step": 140
+    },
+    {
+      "epoch": 0.07721014026508814,
+      "grad_norm": 0.2170065939426422,
+      "learning_rate": 7.660668380462726e-06,
+      "loss": 0.0518,
+      "step": 150
+    },
+    {
+      "epoch": 0.08235748294942735,
+      "grad_norm": 0.20486794412136078,
+      "learning_rate": 8.174807197943445e-06,
+      "loss": 0.0506,
+      "step": 160
+    },
+    {
+      "epoch": 0.08750482563376656,
+      "grad_norm": 0.1600036472082138,
+      "learning_rate": 8.688946015424165e-06,
+      "loss": 0.0505,
+      "step": 170
+    },
+    {
+      "epoch": 0.09265216831810578,
+      "grad_norm": 0.18623106181621552,
+      "learning_rate": 9.203084832904885e-06,
+      "loss": 0.0454,
+      "step": 180
+    },
+    {
+      "epoch": 0.097799511002445,
+      "grad_norm": 0.16273102164268494,
+      "learning_rate": 9.717223650385606e-06,
+      "loss": 0.0443,
+      "step": 190
+    },
+    {
+      "epoch": 0.1029468536867842,
+      "grad_norm": 0.2154380977153778,
+      "learning_rate": 1.0231362467866324e-05,
+      "loss": 0.0421,
+      "step": 200
+    },
+    {
+      "epoch": 0.10809419637112341,
+      "grad_norm": 0.15672603249549866,
+      "learning_rate": 1.0745501285347045e-05,
+      "loss": 0.0413,
+      "step": 210
+    },
+    {
+      "epoch": 0.11324153905546262,
+      "grad_norm": 0.16729186475276947,
+      "learning_rate": 1.1259640102827764e-05,
+      "loss": 0.04,
+      "step": 220
+    },
+    {
+      "epoch": 0.11838888173980183,
+      "grad_norm": 0.17069196701049805,
+      "learning_rate": 1.1773778920308486e-05,
+      "loss": 0.04,
+      "step": 230
+    },
+    {
+      "epoch": 0.12353622442414104,
+      "grad_norm": 0.18878957629203796,
+      "learning_rate": 1.2287917737789203e-05,
+      "loss": 0.0379,
+      "step": 240
+    },
+    {
+      "epoch": 0.12868356710848025,
+      "grad_norm": 0.17584006488323212,
+      "learning_rate": 1.2802056555269925e-05,
+      "loss": 0.0361,
+      "step": 250
+    },
+    {
+      "epoch": 0.13383090979281945,
+      "grad_norm": 0.19568656384944916,
+      "learning_rate": 1.3316195372750644e-05,
+      "loss": 0.0348,
+      "step": 260
+    },
+    {
+      "epoch": 0.13897825247715867,
+      "grad_norm": 0.17687125504016876,
+      "learning_rate": 1.3830334190231362e-05,
+      "loss": 0.0344,
+      "step": 270
+    },
+    {
+      "epoch": 0.14412559516149787,
+      "grad_norm": 0.18790054321289062,
+      "learning_rate": 1.4344473007712083e-05,
+      "loss": 0.0334,
+      "step": 280
+    },
+    {
+      "epoch": 0.1492729378458371,
+      "grad_norm": 0.174034982919693,
+      "learning_rate": 1.4858611825192803e-05,
+      "loss": 0.0343,
+      "step": 290
+    },
+    {
+      "epoch": 0.15442028053017628,
+      "grad_norm": 0.17310510575771332,
+      "learning_rate": 1.5372750642673522e-05,
+      "loss": 0.033,
+      "step": 300
+    },
+    {
+      "epoch": 0.1595676232145155,
+      "grad_norm": 0.16747431457042694,
+      "learning_rate": 1.5886889460154242e-05,
+      "loss": 0.0321,
+      "step": 310
+    },
+    {
+      "epoch": 0.1647149658988547,
+      "grad_norm": 0.19836604595184326,
+      "learning_rate": 1.640102827763496e-05,
+      "loss": 0.0321,
+      "step": 320
+    },
+    {
+      "epoch": 0.16986230858319393,
+      "grad_norm": 0.16714957356452942,
+      "learning_rate": 1.6915167095115684e-05,
+      "loss": 0.0314,
+      "step": 330
+    },
+    {
+      "epoch": 0.17500965126753312,
+      "grad_norm": 0.15047457814216614,
+      "learning_rate": 1.74293059125964e-05,
+      "loss": 0.03,
+      "step": 340
+    },
+    {
+      "epoch": 0.18015699395187235,
+      "grad_norm": 0.153594508767128,
+      "learning_rate": 1.7943444730077123e-05,
+      "loss": 0.0307,
+      "step": 350
+    },
+    {
+      "epoch": 0.18530433663621157,
+      "grad_norm": 0.11620394140481949,
+      "learning_rate": 1.8457583547557843e-05,
+      "loss": 0.0288,
+      "step": 360
+    },
+    {
+      "epoch": 0.19045167932055077,
+      "grad_norm": 0.19321905076503754,
+      "learning_rate": 1.8971722365038563e-05,
+      "loss": 0.0295,
+      "step": 370
+    },
+    {
+      "epoch": 0.19559902200489,
+      "grad_norm": 0.15826652944087982,
+      "learning_rate": 1.9485861182519282e-05,
+      "loss": 0.0288,
+      "step": 380
+    },
+    {
+      "epoch": 0.20074636468922918,
+      "grad_norm": 0.1474819928407669,
+      "learning_rate": 2e-05,
+      "loss": 0.0287,
+      "step": 390
+    },
+    {
+      "epoch": 0.2058937073735684,
+      "grad_norm": 0.15279045701026917,
+      "learning_rate": 1.999959647024453e-05,
+      "loss": 0.0286,
+      "step": 400
+    },
+    {
+      "epoch": 0.2110410500579076,
+      "grad_norm": 0.1290847510099411,
+      "learning_rate": 1.999838591354537e-05,
+      "loss": 0.0269,
+      "step": 410
+    },
+    {
+      "epoch": 0.21618839274224683,
+      "grad_norm": 0.14468011260032654,
+      "learning_rate": 1.9996368427601652e-05,
+      "loss": 0.0272,
+      "step": 420
+    },
+    {
+      "epoch": 0.22133573542658602,
+      "grad_norm": 0.15097224712371826,
+      "learning_rate": 1.9993544175236498e-05,
+      "loss": 0.0263,
+      "step": 430
+    },
+    {
+      "epoch": 0.22648307811092525,
+      "grad_norm": 0.14941558241844177,
+      "learning_rate": 1.998991338438388e-05,
+      "loss": 0.0265,
+      "step": 440
+    },
+    {
+      "epoch": 0.23163042079526444,
+      "grad_norm": 0.12265824526548386,
+      "learning_rate": 1.9985476348070224e-05,
+      "loss": 0.0253,
+      "step": 450
+    },
+    {
+      "epoch": 0.23677776347960366,
+      "grad_norm": 0.13813571631908417,
+      "learning_rate": 1.9980233424390773e-05,
+      "loss": 0.0259,
+      "step": 460
+    },
+    {
+      "epoch": 0.24192510616394286,
+      "grad_norm": 0.13315853476524353,
+      "learning_rate": 1.9974185036480664e-05,
+      "loss": 0.025,
+      "step": 470
+    },
+    {
+      "epoch": 0.24707244884828208,
+      "grad_norm": 0.10484486818313599,
+      "learning_rate": 1.9967331672480798e-05,
+      "loss": 0.0253,
+      "step": 480
+    },
+    {
+      "epoch": 0.2522197915326213,
+      "grad_norm": 0.10516864061355591,
+      "learning_rate": 1.995967388549843e-05,
+      "loss": 0.0236,
+      "step": 490
+    },
+    {
+      "epoch": 0.2573671342169605,
+      "grad_norm": 0.13143527507781982,
+      "learning_rate": 1.9951212293562547e-05,
+      "loss": 0.0247,
+      "step": 500
+    },
+    {
+      "epoch": 0.2625144769012997,
+      "grad_norm": 0.11734788864850998,
+      "learning_rate": 1.994194757957397e-05,
+      "loss": 0.0239,
+      "step": 510
+    },
+    {
+      "epoch": 0.2676618195856389,
+      "grad_norm": 0.10932901501655579,
+      "learning_rate": 1.9931880491250263e-05,
+      "loss": 0.0237,
+      "step": 520
+    },
+    {
+      "epoch": 0.27280916226997814,
+      "grad_norm": 0.10849976539611816,
+      "learning_rate": 1.992101184106535e-05,
+      "loss": 0.0238,
+      "step": 530
+    },
+    {
+      "epoch": 0.27795650495431734,
+      "grad_norm": 0.10729847103357315,
+      "learning_rate": 1.990934250618399e-05,
+      "loss": 0.0247,
+      "step": 540
+    },
+    {
+      "epoch": 0.28310384763865654,
+      "grad_norm": 0.10003025084733963,
+      "learning_rate": 1.989687342839095e-05,
+      "loss": 0.0226,
+      "step": 550
+    },
+    {
+      "epoch": 0.28825119032299573,
+      "grad_norm": 0.1534651517868042,
+      "learning_rate": 1.9883605614015014e-05,
+      "loss": 0.0231,
+      "step": 560
+    },
+    {
+      "epoch": 0.293398533007335,
+      "grad_norm": 0.11108533293008804,
+      "learning_rate": 1.986954013384776e-05,
+      "loss": 0.0223,
+      "step": 570
+    },
+    {
+      "epoch": 0.2985458756916742,
+      "grad_norm": 0.11034880578517914,
+      "learning_rate": 1.9854678123057144e-05,
+      "loss": 0.0231,
+      "step": 580
+    },
+    {
+      "epoch": 0.3036932183760134,
+      "grad_norm": 0.1124214380979538,
+      "learning_rate": 1.9839020781095873e-05,
+      "loss": 0.0218,
+      "step": 590
+    },
+    {
+      "epoch": 0.30884056106035257,
+      "grad_norm": 0.10357758402824402,
+      "learning_rate": 1.9822569371604637e-05,
+      "loss": 0.0223,
+      "step": 600
+    },
+    {
+      "epoch": 0.3139879037446918,
+      "grad_norm": 0.10687187314033508,
+      "learning_rate": 1.9805325222310072e-05,
+      "loss": 0.0207,
+      "step": 610
+    },
+    {
+      "epoch": 0.319135246429031,
+      "grad_norm": 0.11198052763938904,
+      "learning_rate": 1.9787289724917657e-05,
+      "loss": 0.0217,
+      "step": 620
+    },
+    {
+      "epoch": 0.3242825891133702,
+      "grad_norm": 0.10764071345329285,
+      "learning_rate": 1.9768464334999352e-05,
+      "loss": 0.0203,
+      "step": 630
+    },
+    {
+      "epoch": 0.3294299317977094,
+      "grad_norm": 0.11469001322984695,
+      "learning_rate": 1.974885057187617e-05,
+      "loss": 0.0209,
+      "step": 640
+    },
+    {
+      "epoch": 0.33457727448204866,
+      "grad_norm": 0.10319452732801437,
+      "learning_rate": 1.9728450018495506e-05,
+      "loss": 0.0211,
+      "step": 650
+    },
+    {
+      "epoch": 0.33972461716638785,
+      "grad_norm": 0.1113303154706955,
+      "learning_rate": 1.9707264321303427e-05,
+      "loss": 0.021,
+      "step": 660
+    },
+    {
+      "epoch": 0.34487195985072705,
+      "grad_norm": 0.10461831837892532,
+      "learning_rate": 1.9685295190111777e-05,
+      "loss": 0.0204,
+      "step": 670
+    },
+    {
+      "epoch": 0.35001930253506625,
+      "grad_norm": 0.11531244963407516,
+      "learning_rate": 1.9662544397960182e-05,
+      "loss": 0.0202,
+      "step": 680
+    },
+    {
+      "epoch": 0.3551666452194055,
+      "grad_norm": 0.09318046271800995,
+      "learning_rate": 1.9639013780972958e-05,
+      "loss": 0.0205,
+      "step": 690
+    },
+    {
+      "epoch": 0.3603139879037447,
+      "grad_norm": 0.10880846530199051,
+      "learning_rate": 1.961470523821093e-05,
+      "loss": 0.0206,
+      "step": 700
+    },
+    {
+      "epoch": 0.3654613305880839,
+      "grad_norm": 0.1059744656085968,
+      "learning_rate": 1.9589620731518166e-05,
+      "loss": 0.02,
+      "step": 710
+    },
+    {
+      "epoch": 0.37060867327242314,
+      "grad_norm": 0.09044791013002396,
+      "learning_rate": 1.956376228536363e-05,
+      "loss": 0.0197,
+      "step": 720
+    },
+    {
+      "epoch": 0.37575601595676233,
+      "grad_norm": 0.08193252235651016,
+      "learning_rate": 1.953713198667782e-05,
+      "loss": 0.0203,
+      "step": 730
+    },
+    {
+      "epoch": 0.38090335864110153,
+      "grad_norm": 0.10840420424938202,
+      "learning_rate": 1.950973198468431e-05,
+      "loss": 0.0196,
+      "step": 740
+    },
+    {
+      "epoch": 0.3860507013254407,
+      "grad_norm": 0.10423741489648819,
+      "learning_rate": 1.9481564490726327e-05,
+      "loss": 0.0202,
+      "step": 750
+    },
+    {
+      "epoch": 0.39119804400978,
+      "grad_norm": 0.09768087416887283,
+      "learning_rate": 1.9452631778088262e-05,
+      "loss": 0.0192,
+      "step": 760
+    },
+    {
+      "epoch": 0.3963453866941192,
+      "grad_norm": 0.09175275266170502,
+      "learning_rate": 1.94229361818122e-05,
+      "loss": 0.0193,
+      "step": 770
+    },
+    {
+      "epoch": 0.40149272937845837,
+      "grad_norm": 0.10638079047203064,
+      "learning_rate": 1.9392480098509488e-05,
+      "loss": 0.0193,
+      "step": 780
+    },
+    {
+      "epoch": 0.40664007206279756,
+      "grad_norm": 0.0991005077958107,
+      "learning_rate": 1.9361265986167292e-05,
+      "loss": 0.0195,
+      "step": 790
+    },
+    {
+      "epoch": 0.4117874147471368,
+      "grad_norm": 0.09562570601701736,
+      "learning_rate": 1.9329296363950237e-05,
+      "loss": 0.0192,
+      "step": 800
+    },
+    {
+      "epoch": 0.416934757431476,
+      "grad_norm": 0.08249165862798691,
+      "learning_rate": 1.929657381199709e-05,
+      "loss": 0.0194,
+      "step": 810
+    },
+    {
+      "epoch": 0.4220821001158152,
+      "grad_norm": 0.08570217341184616,
+      "learning_rate": 1.9263100971212533e-05,
+      "loss": 0.0195,
+      "step": 820
+    },
+    {
+      "epoch": 0.4272294428001544,
+      "grad_norm": 0.09742734581232071,
+      "learning_rate": 1.922888054305401e-05,
+      "loss": 0.0192,
+      "step": 830
+    },
+    {
+      "epoch": 0.43237678548449365,
+      "grad_norm": 0.0958111509680748,
+      "learning_rate": 1.9193915289313726e-05,
+      "loss": 0.0186,
+      "step": 840
+    },
+    {
+      "epoch": 0.43752412816883285,
+      "grad_norm": 0.10406336188316345,
+      "learning_rate": 1.9158208031895738e-05,
+      "loss": 0.0182,
+      "step": 850
+    },
+    {
+      "epoch": 0.44267147085317204,
+      "grad_norm": 0.09037433564662933,
+      "learning_rate": 1.9121761652588217e-05,
+      "loss": 0.0188,
+      "step": 860
+    },
+    {
+      "epoch": 0.44781881353751124,
+      "grad_norm": 0.09204918146133423,
+      "learning_rate": 1.9084579092830873e-05,
+      "loss": 0.0182,
+      "step": 870
+    },
+    {
+      "epoch": 0.4529661562218505,
+      "grad_norm": 0.07945096492767334,
+      "learning_rate": 1.904666335347755e-05,
+      "loss": 0.017,
+      "step": 880
+    },
+    {
+      "epoch": 0.4581134989061897,
+      "grad_norm": 2.817049980163574,
+      "learning_rate": 1.900801749455406e-05,
+      "loss": 0.0223,
+      "step": 890
+    },
+    {
+      "epoch": 0.4632608415905289,
+      "grad_norm": 1.0506891012191772,
+      "learning_rate": 1.8968644635011192e-05,
+      "loss": 0.0434,
+      "step": 900
+    },
+    {
+      "epoch": 0.4684081842748681,
+      "grad_norm": 0.23307710886001587,
+      "learning_rate": 1.8928547952473037e-05,
+      "loss": 0.0308,
+      "step": 910
+    },
+    {
+      "epoch": 0.47355552695920733,
+      "grad_norm": 0.11467685550451279,
+      "learning_rate": 1.8887730682980484e-05,
+      "loss": 0.0234,
+      "step": 920
+    },
+    {
+      "epoch": 0.4787028696435465,
+      "grad_norm": 0.11315492540597916,
+      "learning_rate": 1.8846196120730095e-05,
+      "loss": 0.0222,
+      "step": 930
+    },
+    {
+      "epoch": 0.4838502123278857,
+      "grad_norm": 0.08601980656385422,
+      "learning_rate": 1.8803947617808217e-05,
+      "loss": 0.0219,
+      "step": 940
+    },
+    {
+      "epoch": 0.4889975550122249,
+      "grad_norm": 0.08733567595481873,
+      "learning_rate": 1.8760988583920457e-05,
+      "loss": 0.0197,
+      "step": 950
+    },
+    {
+      "epoch": 0.49414489769656417,
+      "grad_norm": 0.0972953736782074,
+      "learning_rate": 1.8717322486116513e-05,
+      "loss": 0.019,
+      "step": 960
+    },
+    {
+      "epoch": 0.49929224038090336,
+      "grad_norm": 0.08496998995542526,
+      "learning_rate": 1.867295284851033e-05,
+      "loss": 0.0186,
+      "step": 970
+    },
+    {
+      "epoch": 0.5044395830652426,
+      "grad_norm": 0.09595049172639847,
+      "learning_rate": 1.8627883251995712e-05,
+      "loss": 0.019,
+      "step": 980
+    },
+    {
+      "epoch": 0.5095869257495818,
+      "grad_norm": 0.10503372550010681,
+      "learning_rate": 1.858211733395731e-05,
+      "loss": 0.0175,
+      "step": 990
+    },
+    {
+      "epoch": 0.514734268433921,
+      "grad_norm": 0.07223546504974365,
+      "learning_rate": 1.8535658787977076e-05,
+      "loss": 0.0187,
+      "step": 1000
+    },
+    {
+      "epoch": 0.5198816111182601,
+      "grad_norm": 0.07786504924297333,
+      "learning_rate": 1.848851136353614e-05,
+      "loss": 0.0176,
+      "step": 1010
+    },
+    {
+      "epoch": 0.5250289538025994,
+      "grad_norm": 0.08368386328220367,
+      "learning_rate": 1.8440678865712237e-05,
+      "loss": 0.0171,
+      "step": 1020
+    },
+    {
+      "epoch": 0.5301762964869386,
+      "grad_norm": 0.08910104632377625,
+      "learning_rate": 1.8392165154872596e-05,
+      "loss": 0.0173,
+      "step": 1030
+    },
+    {
+      "epoch": 0.5353236391712778,
+      "grad_norm": 0.07958846539258957,
+      "learning_rate": 1.8342974146362397e-05,
+      "loss": 0.0171,
+      "step": 1040
+    },
+    {
+      "epoch": 0.540470981855617,
+      "grad_norm": 0.08344627916812897,
+      "learning_rate": 1.8293109810188756e-05,
+      "loss": 0.0179,
+      "step": 1050
+    },
+    {
+      "epoch": 0.5456183245399563,
+      "grad_norm": 0.09093711525201797,
+      "learning_rate": 1.8242576170700356e-05,
+      "loss": 0.0182,
+      "step": 1060
+    },
+    {
+      "epoch": 0.5507656672242954,
+      "grad_norm": 0.07439770549535751,
+      "learning_rate": 1.8191377306262635e-05,
+      "loss": 0.0177,
+      "step": 1070
+    },
+    {
+      "epoch": 0.5559130099086347,
+      "grad_norm": 0.07276415079832077,
+      "learning_rate": 1.813951734892864e-05,
+      "loss": 0.017,
+      "step": 1080
+    },
+    {
+      "epoch": 0.5610603525929739,
+      "grad_norm": 0.07302267104387283,
+      "learning_rate": 1.808700048410555e-05,
+      "loss": 0.0173,
+      "step": 1090
+    },
+    {
+      "epoch": 0.5662076952773131,
+      "grad_norm": 0.07959124445915222,
+      "learning_rate": 1.80338309502169e-05,
+      "loss": 0.0171,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5713550379616523,
+      "grad_norm": 0.07042378187179565,
+      "learning_rate": 1.798001303836048e-05,
+      "loss": 0.0177,
+      "step": 1110
+    },
+    {
+      "epoch": 0.5765023806459915,
+      "grad_norm": 0.07214156538248062,
+      "learning_rate": 1.792555109196205e-05,
+      "loss": 0.0176,
+      "step": 1120
+    },
+    {
+      "epoch": 0.5816497233303307,
+      "grad_norm": 0.08310863375663757,
+      "learning_rate": 1.7870449506424802e-05,
+      "loss": 0.0168,
+      "step": 1130
+    },
+    {
+      "epoch": 0.58679706601467,
+      "grad_norm": 0.07772408425807953,
+      "learning_rate": 1.7814712728774602e-05,
+      "loss": 0.0161,
+      "step": 1140
+    },
+    {
+      "epoch": 0.5919444086990091,
+      "grad_norm": 0.07578866183757782,
+      "learning_rate": 1.7758345257301097e-05,
+      "loss": 0.0163,
+      "step": 1150
+    },
+    {
+      "epoch": 0.5970917513833484,
+      "grad_norm": 0.07637569308280945,
+      "learning_rate": 1.770135164119468e-05,
+      "loss": 0.017,
+      "step": 1160
+    },
+    {
+      "epoch": 0.6022390940676876,
+      "grad_norm": 0.07771653681993484,
+      "learning_rate": 1.7643736480179353e-05,
+      "loss": 0.0161,
+      "step": 1170
+    },
+    {
+      "epoch": 0.6073864367520267,
+      "grad_norm": 0.07540572434663773,
+      "learning_rate": 1.7585504424141483e-05,
+      "loss": 0.0162,
+      "step": 1180
+    },
+    {
+      "epoch": 0.612533779436366,
+      "grad_norm": 0.08060440421104431,
+      "learning_rate": 1.752666017275453e-05,
+      "loss": 0.0167,
+      "step": 1190
+    },
+    {
+      "epoch": 0.6176811221207051,
+      "grad_norm": 0.08262394368648529,
+      "learning_rate": 1.7467208475099777e-05,
+      "loss": 0.0164,
+      "step": 1200
+    },
+    {
+      "epoch": 0.6228284648050444,
+      "grad_norm": 0.09898388385772705,
+      "learning_rate": 1.740715412928302e-05,
+      "loss": 0.0161,
+      "step": 1210
+    },
+    {
+      "epoch": 0.6279758074893836,
+      "grad_norm": 0.08121711760759354,
+      "learning_rate": 1.734650198204736e-05,
+      "loss": 0.0158,
+      "step": 1220
+    },
+    {
+      "epoch": 0.6331231501737228,
+      "grad_norm": 0.08025297522544861,
+      "learning_rate": 1.7285256928382024e-05,
+      "loss": 0.0156,
+      "step": 1230
+    },
+    {
+      "epoch": 0.638270492858062,
+      "grad_norm": 0.07479096949100494,
+      "learning_rate": 1.7223423911127315e-05,
+      "loss": 0.0176,
+      "step": 1240
+    },
+    {
+      "epoch": 0.6434178355424013,
+      "grad_norm": 0.08772420138120651,
+      "learning_rate": 1.7161007920575705e-05,
+      "loss": 0.0156,
+      "step": 1250
+    },
+    {
+      "epoch": 0.6485651782267404,
+      "grad_norm": 0.07971519976854324,
+      "learning_rate": 1.709801399406907e-05,
+      "loss": 0.0154,
+      "step": 1260
+    },
+    {
+      "epoch": 0.6537125209110797,
+      "grad_norm": 0.07564999163150787,
+      "learning_rate": 1.7034447215592168e-05,
+      "loss": 0.0154,
+      "step": 1270
+    },
+    {
+      "epoch": 0.6588598635954188,
+      "grad_norm": 0.07267491519451141,
+      "learning_rate": 1.6970312715362304e-05,
+      "loss": 0.0159,
+      "step": 1280
+    },
+    {
+      "epoch": 0.6640072062797581,
+      "grad_norm": 0.07202299684286118,
+      "learning_rate": 1.6905615669415326e-05,
+      "loss": 0.0164,
+      "step": 1290
+    },
+    {
+      "epoch": 0.6691545489640973,
+      "grad_norm": 0.07982448488473892,
+      "learning_rate": 1.684036129918786e-05,
+      "loss": 0.0155,
+      "step": 1300
+    },
+    {
+      "epoch": 0.6743018916484365,
+      "grad_norm": 0.07582129538059235,
+      "learning_rate": 1.6774554871095918e-05,
+      "loss": 0.0153,
+      "step": 1310
+    },
+    {
+      "epoch": 0.6794492343327757,
+      "grad_norm": 0.08038461208343506,
+      "learning_rate": 1.6708201696109857e-05,
+      "loss": 0.0158,
+      "step": 1320
+    },
+    {
+      "epoch": 0.684596577017115,
+      "grad_norm": 0.07446227967739105,
+      "learning_rate": 1.6641307129325785e-05,
+      "loss": 0.0164,
+      "step": 1330
+    },
+    {
+      "epoch": 0.6897439197014541,
+      "grad_norm": 0.0867164209485054,
+      "learning_rate": 1.657387656953333e-05,
+      "loss": 0.016,
+      "step": 1340
+    },
+    {
+      "epoch": 0.6948912623857934,
+      "grad_norm": 0.07672440260648727,
+      "learning_rate": 1.6505915458779958e-05,
+      "loss": 0.0153,
+      "step": 1350
+    },
+    {
+      "epoch": 0.7000386050701325,
+      "grad_norm": 0.07314340770244598,
+      "learning_rate": 1.6437429281931744e-05,
+      "loss": 0.0145,
+      "step": 1360
+    },
+    {
+      "epoch": 0.7051859477544717,
+      "grad_norm": 0.07686632126569748,
+      "learning_rate": 1.636842356623073e-05,
+      "loss": 0.0152,
+      "step": 1370
+    },
+    {
+      "epoch": 0.710333290438811,
+      "grad_norm": 0.07814227044582367,
+      "learning_rate": 1.6298903880848836e-05,
+      "loss": 0.0155,
+      "step": 1380
+    },
+    {
+      "epoch": 0.7154806331231501,
+      "grad_norm": 0.07057350873947144,
+      "learning_rate": 1.6228875836438386e-05,
+      "loss": 0.0152,
+      "step": 1390
+    },
+    {
+      "epoch": 0.7206279758074894,
+      "grad_norm": 0.06801946461200714,
+      "learning_rate": 1.615834508467931e-05,
+      "loss": 0.0148,
+      "step": 1400
+    },
+    {
+      "epoch": 0.7257753184918286,
+      "grad_norm": 0.08162187039852142,
+      "learning_rate": 1.608731731782301e-05,
+      "loss": 0.0156,
+      "step": 1410
+    },
+    {
+      "epoch": 0.7309226611761678,
+      "grad_norm": 0.09547685086727142,
+      "learning_rate": 1.6015798268232964e-05,
+      "loss": 0.0158,
+      "step": 1420
+    },
+    {
+      "epoch": 0.736070003860507,
+      "grad_norm": 0.07891667634248734,
+      "learning_rate": 1.5943793707922086e-05,
+      "loss": 0.0144,
+      "step": 1430
+    },
+    {
+      "epoch": 0.7412173465448463,
+      "grad_norm": 0.07318150252103806,
+      "learning_rate": 1.5871309448086903e-05,
+      "loss": 0.0158,
+      "step": 1440
+    },
+    {
+      "epoch": 0.7463646892291854,
+      "grad_norm": 0.06992822885513306,
+      "learning_rate": 1.5798351338638548e-05,
+      "loss": 0.0148,
+      "step": 1450
+    },
+    {
+      "epoch": 0.7515120319135247,
+      "grad_norm": 0.0747927725315094,
+      "learning_rate": 1.5724925267730625e-05,
+      "loss": 0.0156,
+      "step": 1460
+    },
+    {
+      "epoch": 0.7566593745978638,
+      "grad_norm": 0.06734950840473175,
+      "learning_rate": 1.5651037161284032e-05,
+      "loss": 0.0155,
+      "step": 1470
+    },
+    {
+      "epoch": 0.7618067172822031,
+      "grad_norm": 0.06667228043079376,
+      "learning_rate": 1.5576692982508665e-05,
+      "loss": 0.0149,
+      "step": 1480
+    },
+    {
+      "epoch": 0.7669540599665423,
+      "grad_norm": 0.07065103948116302,
+      "learning_rate": 1.550189873142219e-05,
+      "loss": 0.0149,
+      "step": 1490
+    },
+    {
+      "epoch": 0.7721014026508815,
+      "grad_norm": 0.07318428158760071,
+      "learning_rate": 1.542666044436577e-05,
+      "loss": 0.0147,
+      "step": 1500
+    },
+    {
+      "epoch": 0.7772487453352207,
+      "grad_norm": 0.06767120212316513,
+      "learning_rate": 1.5350984193516917e-05,
+      "loss": 0.0152,
+      "step": 1510
+    },
+    {
+      "epoch": 0.78239608801956,
+      "grad_norm": 0.08945749700069427,
+      "learning_rate": 1.5274876086399444e-05,
+      "loss": 0.0147,
+      "step": 1520
+    },
+    {
+      "epoch": 0.7875434307038991,
+      "grad_norm": 0.07100491225719452,
+      "learning_rate": 1.5198342265390511e-05,
+      "loss": 0.0144,
+      "step": 1530
+    },
+    {
+      "epoch": 0.7926907733882383,
+      "grad_norm": 0.06846942007541656,
+      "learning_rate": 1.512138890722494e-05,
+      "loss": 0.0148,
+      "step": 1540
+    },
+    {
+      "epoch": 0.7978381160725775,
+      "grad_norm": 0.07432665675878525,
+      "learning_rate": 1.5044022222496684e-05,
+      "loss": 0.0153,
+      "step": 1550
+    },
+    {
+      "epoch": 0.8029854587569167,
+      "grad_norm": 0.06611491739749908,
+      "learning_rate": 1.4966248455157622e-05,
+      "loss": 0.0143,
+      "step": 1560
+    },
+    {
+      "epoch": 0.808132801441256,
+      "grad_norm": 0.07415001094341278,
+      "learning_rate": 1.4888073882013616e-05,
+      "loss": 0.0143,
+      "step": 1570
+    },
+    {
+      "epoch": 0.8132801441255951,
+      "grad_norm": 0.06780339032411575,
+      "learning_rate": 1.480950481221795e-05,
+      "loss": 0.0145,
+      "step": 1580
+    },
+    {
+      "epoch": 0.8184274868099344,
+      "grad_norm": 0.07141536474227905,
+      "learning_rate": 1.4730547586762114e-05,
+      "loss": 0.0139,
+      "step": 1590
+    },
+    {
+      "epoch": 0.8235748294942736,
+      "grad_norm": 0.07404403388500214,
+      "learning_rate": 1.4651208577964099e-05,
+      "loss": 0.0149,
+      "step": 1600
+    },
+    {
+      "epoch": 0.8287221721786128,
+      "grad_norm": 0.07202573120594025,
+      "learning_rate": 1.4571494188954058e-05,
+      "loss": 0.0149,
+      "step": 1610
+    },
+    {
+      "epoch": 0.833869514862952,
+      "grad_norm": 0.055626142770051956,
+      "learning_rate": 1.4491410853157577e-05,
+      "loss": 0.014,
+      "step": 1620
+    },
+    {
+      "epoch": 0.8390168575472912,
+      "grad_norm": 0.06371673196554184,
+      "learning_rate": 1.4410965033776439e-05,
+      "loss": 0.0146,
+      "step": 1630
+    },
+    {
+      "epoch": 0.8441642002316304,
+      "grad_norm": 0.0604044534265995,
+      "learning_rate": 1.4330163223267005e-05,
+      "loss": 0.0147,
+      "step": 1640
+    },
+    {
+      "epoch": 0.8493115429159697,
+      "grad_norm": 0.07098476588726044,
+      "learning_rate": 1.4249011942816245e-05,
+      "loss": 0.0148,
+      "step": 1650
+    },
+    {
+      "epoch": 0.8544588856003088,
+      "grad_norm": 0.06837648898363113,
+      "learning_rate": 1.416751774181543e-05,
+      "loss": 0.0144,
+      "step": 1660
+    },
+    {
+      "epoch": 0.859606228284648,
+      "grad_norm": 0.06875620782375336,
+      "learning_rate": 1.4085687197331559e-05,
+      "loss": 0.0143,
+      "step": 1670
+    },
+    {
+      "epoch": 0.8647535709689873,
+      "grad_norm": 0.0676610991358757,
+      "learning_rate": 1.4003526913576555e-05,
+      "loss": 0.0152,
+      "step": 1680
+    },
+    {
+      "epoch": 0.8699009136533264,
+      "grad_norm": 0.06219542399048805,
+      "learning_rate": 1.392104352137426e-05,
+      "loss": 0.0137,
+      "step": 1690
+    },
+    {
+      "epoch": 0.8750482563376657,
+      "grad_norm": 0.06112512946128845,
+      "learning_rate": 1.3838243677625292e-05,
+      "loss": 0.0145,
+      "step": 1700
+    },
+    {
+      "epoch": 0.8801955990220048,
+      "grad_norm": 0.07453671842813492,
+      "learning_rate": 1.3755134064769788e-05,
+      "loss": 0.0144,
+      "step": 1710
+    },
+    {
+      "epoch": 0.8853429417063441,
+      "grad_norm": 0.07224484533071518,
+      "learning_rate": 1.36717213902481e-05,
+      "loss": 0.0138,
+      "step": 1720
+    },
+    {
+      "epoch": 0.8904902843906833,
+      "grad_norm": 0.06484848260879517,
+      "learning_rate": 1.3588012385959464e-05,
+      "loss": 0.0148,
+      "step": 1730
+    },
+    {
+      "epoch": 0.8956376270750225,
+      "grad_norm": 0.06340837478637695,
+      "learning_rate": 1.350401380771868e-05,
+      "loss": 0.014,
+      "step": 1740
+    },
+    {
+      "epoch": 0.9007849697593617,
+      "grad_norm": 0.06442750245332718,
+      "learning_rate": 1.3419732434710895e-05,
+      "loss": 0.0134,
+      "step": 1750
+    },
+    {
+      "epoch": 0.905932312443701,
+      "grad_norm": 0.06431768089532852,
+      "learning_rate": 1.333517506894448e-05,
+      "loss": 0.0144,
+      "step": 1760
+    },
+    {
+      "epoch": 0.9110796551280401,
+      "grad_norm": 0.06857169419527054,
+      "learning_rate": 1.3250348534702063e-05,
+      "loss": 0.0141,
+      "step": 1770
+    },
+    {
+      "epoch": 0.9162269978123794,
+      "grad_norm": 0.07051096856594086,
+      "learning_rate": 1.316525967798976e-05,
+      "loss": 0.0141,
+      "step": 1780
+    },
+    {
+      "epoch": 0.9213743404967185,
+      "grad_norm": 0.07205208390951157,
+      "learning_rate": 1.307991536598469e-05,
+      "loss": 0.0146,
+      "step": 1790
+    },
+    {
+      "epoch": 0.9265216831810578,
+      "grad_norm": 0.07224377989768982,
+      "learning_rate": 1.2994322486480716e-05,
+      "loss": 0.0137,
+      "step": 1800
+    },
+    {
+      "epoch": 0.931669025865397,
+      "grad_norm": 0.0663972795009613,
+      "learning_rate": 1.2908487947332588e-05,
+      "loss": 0.0135,
+      "step": 1810
+    },
+    {
+      "epoch": 0.9368163685497362,
+      "grad_norm": 0.06685107946395874,
+      "learning_rate": 1.2822418675898428e-05,
+      "loss": 0.0146,
+      "step": 1820
+    },
+    {
+      "epoch": 0.9419637112340754,
+      "grad_norm": 0.07378830760717392,
+      "learning_rate": 1.273612161848064e-05,
+      "loss": 0.0137,
+      "step": 1830
+    },
+    {
+      "epoch": 0.9471110539184147,
+      "grad_norm": 0.07926513254642487,
+      "learning_rate": 1.2649603739765323e-05,
+      "loss": 0.0134,
+      "step": 1840
+    },
+    {
+      "epoch": 0.9522583966027538,
+      "grad_norm": 0.060047443956136703,
+      "learning_rate": 1.2562872022260168e-05,
+      "loss": 0.0142,
+      "step": 1850
+    },
+    {
+      "epoch": 0.957405739287093,
+      "grad_norm": 0.07245929539203644,
+      "learning_rate": 1.2475933465730923e-05,
+      "loss": 0.0134,
+      "step": 1860
+    },
+    {
+      "epoch": 0.9625530819714323,
+      "grad_norm": 0.06252739578485489,
+      "learning_rate": 1.238879508663648e-05,
+      "loss": 0.0136,
+      "step": 1870
+    },
+    {
+      "epoch": 0.9677004246557714,
+      "grad_norm": 0.07423530519008636,
+      "learning_rate": 1.2301463917562602e-05,
+      "loss": 0.0138,
+      "step": 1880
+    },
+    {
+      "epoch": 0.9728477673401107,
+      "grad_norm": 0.06203414499759674,
+      "learning_rate": 1.2213947006654347e-05,
+      "loss": 0.0136,
+      "step": 1890
+    },
+    {
+      "epoch": 0.9779951100244498,
+      "grad_norm": 0.06526093184947968,
+      "learning_rate": 1.212625141704725e-05,
+      "loss": 0.0144,
+      "step": 1900
+    },
+    {
+      "epoch": 0.9831424527087891,
+      "grad_norm": 0.05829852074384689,
+      "learning_rate": 1.2038384226297272e-05,
+      "loss": 0.0135,
+      "step": 1910
+    },
+    {
+      "epoch": 0.9882897953931283,
+      "grad_norm": 0.0650823563337326,
+      "learning_rate": 1.1950352525809619e-05,
+      "loss": 0.0134,
+      "step": 1920
+    },
+    {
+      "epoch": 0.9934371380774675,
+      "grad_norm": 0.07050953805446625,
+      "learning_rate": 1.18621634202664e-05,
+      "loss": 0.0136,
+      "step": 1930
+    },
+    {
+      "epoch": 0.9985844807618067,
+      "grad_norm": 0.05969216674566269,
+      "learning_rate": 1.1773824027053256e-05,
+      "loss": 0.0135,
+      "step": 1940
+    },
+    {
+      "epoch": 1.0036031398790375,
+      "grad_norm": 0.05986722931265831,
+      "learning_rate": 1.1685341475684935e-05,
+      "loss": 0.0123,
+      "step": 1950
+    },
+    {
+      "epoch": 1.0087504825633766,
+      "grad_norm": 0.06715232133865356,
+      "learning_rate": 1.15967229072299e-05,
+      "loss": 0.0125,
+      "step": 1960
+    },
+    {
+      "epoch": 1.0138978252477158,
+      "grad_norm": 0.05757083743810654,
+      "learning_rate": 1.1507975473734002e-05,
+      "loss": 0.0123,
+      "step": 1970
+    },
+    {
+      "epoch": 1.0190451679320551,
+      "grad_norm": 0.06696327775716782,
+      "learning_rate": 1.141910633764327e-05,
+      "loss": 0.0122,
+      "step": 1980
+    },
+    {
+      "epoch": 1.0241925106163943,
+      "grad_norm": 0.061676956713199615,
+      "learning_rate": 1.1330122671225855e-05,
+      "loss": 0.0125,
+      "step": 1990
+    },
+    {
+      "epoch": 1.0293398533007334,
+      "grad_norm": 0.07469218969345093,
+      "learning_rate": 1.1241031655993188e-05,
+      "loss": 0.0127,
+      "step": 2000
+    },
+    {
+      "epoch": 1.0344871959850728,
+      "grad_norm": 0.06639962643384933,
+      "learning_rate": 1.1151840482120386e-05,
+      "loss": 0.0126,
+      "step": 2010
+    },
+    {
+      "epoch": 1.039634538669412,
+      "grad_norm": 0.07054606825113297,
+      "learning_rate": 1.1062556347865967e-05,
+      "loss": 0.0127,
+      "step": 2020
+    },
+    {
+      "epoch": 1.044781881353751,
+      "grad_norm": 0.06040499359369278,
+      "learning_rate": 1.0973186458990902e-05,
+      "loss": 0.0119,
+      "step": 2030
+    },
+    {
+      "epoch": 1.0499292240380904,
+      "grad_norm": 0.0656294897198677,
+      "learning_rate": 1.0883738028177069e-05,
+      "loss": 0.0126,
+      "step": 2040
+    },
+    {
+      "epoch": 1.0550765667224296,
+      "grad_norm": 0.06062242388725281,
+      "learning_rate": 1.0794218274445155e-05,
+      "loss": 0.0128,
+      "step": 2050
+    },
+    {
+      "epoch": 1.0602239094067687,
+      "grad_norm": 0.06211702898144722,
+      "learning_rate": 1.0704634422572029e-05,
+      "loss": 0.0127,
+      "step": 2060
+    },
+    {
+      "epoch": 1.065371252091108,
+      "grad_norm": 0.05939929559826851,
+      "learning_rate": 1.0614993702507649e-05,
+      "loss": 0.0129,
+      "step": 2070
+    },
+    {
+      "epoch": 1.0705185947754472,
+      "grad_norm": 0.06660640239715576,
+      "learning_rate": 1.0525303348791599e-05,
+      "loss": 0.012,
+      "step": 2080
+    },
+    {
+      "epoch": 1.0756659374597863,
+      "grad_norm": 0.061506547033786774,
+      "learning_rate": 1.0435570599969168e-05,
+      "loss": 0.0126,
+      "step": 2090
+    },
+    {
+      "epoch": 1.0808132801441257,
+      "grad_norm": 0.05894378572702408,
+      "learning_rate": 1.0345802698007198e-05,
+      "loss": 0.0127,
+      "step": 2100
+    },
+    {
+      "epoch": 1.0859606228284648,
+      "grad_norm": 0.06136978417634964,
+      "learning_rate": 1.0256006887709593e-05,
+      "loss": 0.0115,
+      "step": 2110
+    },
+    {
+      "epoch": 1.091107965512804,
+      "grad_norm": 0.0651112049818039,
+      "learning_rate": 1.016619041613263e-05,
+      "loss": 0.0125,
+      "step": 2120
+    },
+    {
+      "epoch": 1.0962553081971431,
+      "grad_norm": 0.06708134710788727,
+      "learning_rate": 1.0076360532000069e-05,
+      "loss": 0.0126,
+      "step": 2130
+    },
+    {
+      "epoch": 1.1014026508814825,
+      "grad_norm": 0.0627651959657669,
+      "learning_rate": 9.986524485118152e-06,
+      "loss": 0.0119,
+      "step": 2140
+    },
+    {
+      "epoch": 1.1065499935658216,
+      "grad_norm": 0.060133691877126694,
+      "learning_rate": 9.896689525790474e-06,
+      "loss": 0.013,
+      "step": 2150
+    },
+    {
+      "epoch": 1.1116973362501608,
+      "grad_norm": 0.05978613346815109,
+      "learning_rate": 9.80686290423288e-06,
+      "loss": 0.0118,
+      "step": 2160
+    },
+    {
+      "epoch": 1.1168446789345001,
+      "grad_norm": 0.0568259134888649,
+      "learning_rate": 9.717051869988287e-06,
+      "loss": 0.0122,
+      "step": 2170
+    },
+    {
+      "epoch": 1.1219920216188393,
+      "grad_norm": 0.07073856145143509,
+      "learning_rate": 9.627263671341638e-06,
+      "loss": 0.012,
+      "step": 2180
+    },
+    {
+      "epoch": 1.1271393643031784,
+      "grad_norm": 0.0650116428732872,
+      "learning_rate": 9.537505554734901e-06,
+      "loss": 0.0126,
+      "step": 2190
+    },
+    {
+      "epoch": 1.1322867069875178,
+      "grad_norm": 0.06662643700838089,
+      "learning_rate": 9.447784764182247e-06,
+      "loss": 0.0121,
+      "step": 2200
+    },
+    {
+      "epoch": 1.137434049671857,
+      "grad_norm": 0.06186177209019661,
+      "learning_rate": 9.358108540685406e-06,
+      "loss": 0.0116,
+      "step": 2210
+    },
+    {
+      "epoch": 1.142581392356196,
+      "grad_norm": 0.06745568662881851,
+      "learning_rate": 9.268484121649289e-06,
+      "loss": 0.0126,
+      "step": 2220
+    },
+    {
+      "epoch": 1.1477287350405354,
+      "grad_norm": 0.0625823512673378,
+      "learning_rate": 9.178918740297877e-06,
+      "loss": 0.0135,
+      "step": 2230
+    },
+    {
+      "epoch": 1.1528760777248745,
+      "grad_norm": 0.05811070650815964,
+      "learning_rate": 9.08941962509045e-06,
+      "loss": 0.0122,
+      "step": 2240
+    },
+    {
+      "epoch": 1.1580234204092137,
+      "grad_norm": 0.06800320744514465,
+      "learning_rate": 8.999993999138227e-06,
+      "loss": 0.0128,
+      "step": 2250
+    },
+    {
+      "epoch": 1.1631707630935528,
+      "grad_norm": 0.061803579330444336,
+      "learning_rate": 8.91064907962141e-06,
+      "loss": 0.0116,
+      "step": 2260
+    },
+    {
+      "epoch": 1.1683181057778922,
+      "grad_norm": 0.056932009756565094,
+      "learning_rate": 8.821392077206695e-06,
+      "loss": 0.0123,
+      "step": 2270
+    },
+    {
+      "epoch": 1.1734654484622313,
+      "grad_norm": 0.0628976821899414,
+      "learning_rate": 8.732230195465353e-06,
+      "loss": 0.0119,
+      "step": 2280
+    },
+    {
+      "epoch": 1.1786127911465707,
+      "grad_norm": 0.0606074184179306,
+      "learning_rate": 8.64317063029186e-06,
+      "loss": 0.0121,
+      "step": 2290
+    },
+    {
+      "epoch": 1.1837601338309098,
+      "grad_norm": 0.06012739986181259,
+      "learning_rate": 8.554220569323117e-06,
+      "loss": 0.0123,
+      "step": 2300
+    },
+    {
+      "epoch": 1.188907476515249,
+      "grad_norm": 0.060469817370176315,
+      "learning_rate": 8.465387191358398e-06,
+      "loss": 0.0121,
+      "step": 2310
+    },
+    {
+      "epoch": 1.1940548191995881,
+      "grad_norm": 0.06011821702122688,
+      "learning_rate": 8.376677665779968e-06,
+      "loss": 0.0119,
+      "step": 2320
+    },
+    {
+      "epoch": 1.1992021618839275,
+      "grad_norm": 0.0587395578622818,
+      "learning_rate": 8.288099151974446e-06,
+      "loss": 0.0121,
+      "step": 2330
+    },
+    {
+      "epoch": 1.2043495045682666,
+      "grad_norm": 0.06356833130121231,
+      "learning_rate": 8.199658798755048e-06,
+      "loss": 0.0123,
+      "step": 2340
+    },
+    {
+      "epoch": 1.2094968472526058,
+      "grad_norm": 0.05952199548482895,
+      "learning_rate": 8.111363743784587e-06,
+      "loss": 0.0123,
+      "step": 2350
+    },
+    {
+      "epoch": 1.2146441899369451,
+      "grad_norm": 0.0687473863363266,
+      "learning_rate": 8.023221112999459e-06,
+      "loss": 0.0117,
+      "step": 2360
+    },
+    {
+      "epoch": 1.2197915326212843,
+      "grad_norm": 0.05651247128844261,
+      "learning_rate": 7.935238020034505e-06,
+      "loss": 0.0107,
+      "step": 2370
+    },
+    {
+      "epoch": 1.2249388753056234,
+      "grad_norm": 0.05863620713353157,
+      "learning_rate": 7.847421565648929e-06,
+      "loss": 0.0119,
+      "step": 2380
+    },
+    {
+      "epoch": 1.2300862179899628,
+      "grad_norm": 0.06163261458277702,
+      "learning_rate": 7.759778837153205e-06,
+      "loss": 0.0125,
+      "step": 2390
+    },
+    {
+      "epoch": 1.235233560674302,
+      "grad_norm": 0.061220042407512665,
+      "learning_rate": 7.672316907837084e-06,
+      "loss": 0.0118,
+      "step": 2400
+    },
+    {
+      "epoch": 1.240380903358641,
+      "grad_norm": 0.06481561809778214,
+      "learning_rate": 7.585042836398763e-06,
+      "loss": 0.0118,
+      "step": 2410
+    },
+    {
+      "epoch": 1.2455282460429804,
+      "grad_norm": 0.06460854411125183,
+      "learning_rate": 7.497963666375185e-06,
+      "loss": 0.0124,
+      "step": 2420
+    },
+    {
+      "epoch": 1.2506755887273195,
+      "grad_norm": 0.0643194317817688,
+      "learning_rate": 7.411086425573584e-06,
+      "loss": 0.0119,
+      "step": 2430
+    },
+    {
+      "epoch": 1.2558229314116587,
+      "grad_norm": 0.0629289299249649,
+      "learning_rate": 7.324418125504305e-06,
+      "loss": 0.0123,
+      "step": 2440
+    },
+    {
+      "epoch": 1.2609702740959978,
+      "grad_norm": 0.06404928117990494,
+      "learning_rate": 7.23796576081494e-06,
+      "loss": 0.0122,
+      "step": 2450
+    },
+    {
+      "epoch": 1.2661176167803372,
+      "grad_norm": 0.06378574669361115,
+      "learning_rate": 7.151736308725802e-06,
+      "loss": 0.0114,
+      "step": 2460
+    },
+    {
+      "epoch": 1.2712649594646763,
+      "grad_norm": 0.06403859704732895,
+      "learning_rate": 7.065736728466832e-06,
+      "loss": 0.0117,
+      "step": 2470
+    },
+    {
+      "epoch": 1.2764123021490157,
+      "grad_norm": 0.06829800456762314,
+      "learning_rate": 6.979973960715958e-06,
+      "loss": 0.0124,
+      "step": 2480
+    },
+    {
+      "epoch": 1.2815596448333548,
+      "grad_norm": 0.059846751391887665,
+      "learning_rate": 6.894454927038907e-06,
+      "loss": 0.0113,
+      "step": 2490
+    },
+    {
+      "epoch": 1.286706987517694,
+      "grad_norm": 0.054608747363090515,
+      "learning_rate": 6.809186529330639e-06,
+      "loss": 0.012,
+      "step": 2500
+    },
+    {
+      "epoch": 1.291854330202033,
+      "grad_norm": 0.06600859761238098,
+      "learning_rate": 6.724175649258287e-06,
+      "loss": 0.0126,
+      "step": 2510
+    },
+    {
+      "epoch": 1.2970016728863725,
+      "grad_norm": 0.05849931016564369,
+      "learning_rate": 6.6394291477057736e-06,
+      "loss": 0.0116,
+      "step": 2520
+    },
+    {
+      "epoch": 1.3021490155707116,
+      "grad_norm": 0.06548711657524109,
+      "learning_rate": 6.554953864220115e-06,
+      "loss": 0.0121,
+      "step": 2530
+    },
+    {
+      "epoch": 1.3072963582550507,
+      "grad_norm": 0.05490358918905258,
+      "learning_rate": 6.4707566164594105e-06,
+      "loss": 0.0116,
+      "step": 2540
+    },
+    {
+      "epoch": 1.3124437009393901,
+      "grad_norm": 0.05758791044354439,
+      "learning_rate": 6.386844199642619e-06,
+      "loss": 0.0117,
+      "step": 2550
+    },
+    {
+      "epoch": 1.3175910436237293,
+      "grad_norm": 0.0644078403711319,
+      "learning_rate": 6.30322338600114e-06,
+      "loss": 0.0124,
+      "step": 2560
+    },
+    {
+      "epoch": 1.3227383863080684,
+      "grad_norm": 0.05672595277428627,
+      "learning_rate": 6.219900924232279e-06,
+      "loss": 0.011,
+      "step": 2570
+    },
+    {
+      "epoch": 1.3278857289924075,
+      "grad_norm": 0.058410342782735825,
+      "learning_rate": 6.136883538954561e-06,
+      "loss": 0.012,
+      "step": 2580
+    },
+    {
+      "epoch": 1.333033071676747,
+      "grad_norm": 0.05841913819313049,
+      "learning_rate": 6.054177930165017e-06,
+      "loss": 0.0111,
+      "step": 2590
+    },
+    {
+      "epoch": 1.338180414361086,
+      "grad_norm": 0.06315062940120697,
+      "learning_rate": 5.971790772698467e-06,
+      "loss": 0.0114,
+      "step": 2600
+    },
+    {
+      "epoch": 1.3433277570454254,
+      "grad_norm": 0.059485290199518204,
+      "learning_rate": 5.889728715688814e-06,
+      "loss": 0.0116,
+      "step": 2610
+    },
+    {
+      "epoch": 1.3484750997297645,
+      "grad_norm": 0.06330478936433792,
+      "learning_rate": 5.807998382032414e-06,
+      "loss": 0.0117,
+      "step": 2620
+    },
+    {
+      "epoch": 1.3536224424141037,
+      "grad_norm": 0.05869928374886513,
+      "learning_rate": 5.726606367853581e-06,
+      "loss": 0.0119,
+      "step": 2630
+    },
+    {
+      "epoch": 1.3587697850984428,
+      "grad_norm": 0.07355443388223648,
+      "learning_rate": 5.645559241972231e-06,
+      "loss": 0.0123,
+      "step": 2640
+    },
+    {
+      "epoch": 1.3639171277827822,
+      "grad_norm": 0.05870656296610832,
+      "learning_rate": 5.56486354537374e-06,
+      "loss": 0.0117,
+      "step": 2650
+    },
+    {
+      "epoch": 1.3690644704671213,
+      "grad_norm": 0.059707384556531906,
+      "learning_rate": 5.484525790681052e-06,
+      "loss": 0.0117,
+      "step": 2660
+    },
+    {
+      "epoch": 1.3742118131514607,
+      "grad_norm": 0.0809202715754509,
+      "learning_rate": 5.404552461629069e-06,
+      "loss": 0.0114,
+      "step": 2670
+    },
+    {
+      "epoch": 1.3793591558357998,
+      "grad_norm": 0.05704690143465996,
+      "learning_rate": 5.324950012541372e-06,
+      "loss": 0.0118,
+      "step": 2680
+    },
+    {
+      "epoch": 1.384506498520139,
+      "grad_norm": 0.06190679967403412,
+      "learning_rate": 5.245724867809326e-06,
+      "loss": 0.0116,
+      "step": 2690
+    },
+    {
+      "epoch": 1.389653841204478,
+      "grad_norm": 0.055938467383384705,
+      "learning_rate": 5.166883421373583e-06,
+      "loss": 0.0116,
+      "step": 2700
+    },
+    {
+      "epoch": 1.3948011838888175,
+      "grad_norm": 0.05830509215593338,
+      "learning_rate": 5.088432036208075e-06,
+      "loss": 0.0112,
+      "step": 2710
+    },
+    {
+      "epoch": 1.3999485265731566,
+      "grad_norm": 0.05540473014116287,
+      "learning_rate": 5.010377043806443e-06,
+      "loss": 0.0102,
+      "step": 2720
+    },
+    {
+      "epoch": 1.4050958692574957,
+      "grad_norm": 0.05529293790459633,
+      "learning_rate": 4.932724743671089e-06,
+      "loss": 0.0113,
+      "step": 2730
+    },
+    {
+      "epoch": 1.410243211941835,
+      "grad_norm": 0.062383200973272324,
+      "learning_rate": 4.8554814028047605e-06,
+      "loss": 0.0114,
+      "step": 2740
+    },
+    {
+      "epoch": 1.4153905546261742,
+      "grad_norm": 0.05733776092529297,
+      "learning_rate": 4.778653255204733e-06,
+      "loss": 0.012,
+      "step": 2750
+    },
+    {
+      "epoch": 1.4205378973105134,
+      "grad_norm": 0.05461714044213295,
+      "learning_rate": 4.702246501359739e-06,
+      "loss": 0.0108,
+      "step": 2760
+    },
+    {
+      "epoch": 1.4256852399948525,
+      "grad_norm": 0.05713033676147461,
+      "learning_rate": 4.626267307749519e-06,
+      "loss": 0.0113,
+      "step": 2770
+    },
+    {
+      "epoch": 1.4308325826791919,
+      "grad_norm": 0.066215381026268,
+      "learning_rate": 4.550721806347146e-06,
+      "loss": 0.0117,
+      "step": 2780
+    },
+    {
+      "epoch": 1.435979925363531,
+      "grad_norm": 0.05536945164203644,
+      "learning_rate": 4.475616094124171e-06,
+      "loss": 0.0114,
+      "step": 2790
+    },
+    {
+      "epoch": 1.4411272680478704,
+      "grad_norm": 0.05297793820500374,
+      "learning_rate": 4.40095623255853e-06,
+      "loss": 0.0111,
+      "step": 2800
+    },
+    {
+      "epoch": 1.4462746107322095,
+      "grad_norm": 0.05898694694042206,
+      "learning_rate": 4.3267482471453574e-06,
+      "loss": 0.0107,
+      "step": 2810
+    },
+    {
+      "epoch": 1.4514219534165487,
+      "grad_norm": 0.05825399234890938,
+      "learning_rate": 4.252998126910696e-06,
+      "loss": 0.01,
+      "step": 2820
+    },
+    {
+      "epoch": 1.4565692961008878,
+      "grad_norm": 0.05541868135333061,
+      "learning_rate": 4.179711823928141e-06,
+      "loss": 0.0104,
+      "step": 2830
+    },
+    {
+      "epoch": 1.4617166387852272,
+      "grad_norm": 0.061555661261081696,
+      "learning_rate": 4.106895252838478e-06,
+      "loss": 0.0108,
+      "step": 2840
+    },
+    {
+      "epoch": 1.4668639814695663,
+      "grad_norm": 0.05560790374875069,
+      "learning_rate": 4.034554290372332e-06,
+      "loss": 0.0113,
+      "step": 2850
+    },
+    {
+      "epoch": 1.4720113241539057,
+      "grad_norm": 0.06007440760731697,
+      "learning_rate": 3.96269477487588e-06,
+      "loss": 0.0106,
+      "step": 2860
+    },
+    {
+      "epoch": 1.4771586668382448,
+      "grad_norm": 0.0683414563536644,
+      "learning_rate": 3.891322505839672e-06,
+      "loss": 0.0112,
+      "step": 2870
+    },
+    {
+      "epoch": 1.482306009522584,
+      "grad_norm": 0.06036698445677757,
+      "learning_rate": 3.820443243430555e-06,
+      "loss": 0.0113,
+      "step": 2880
+    },
+    {
+      "epoch": 1.487453352206923,
+      "grad_norm": 0.060313206166028976,
+      "learning_rate": 3.750062708026814e-06,
+      "loss": 0.0113,
+      "step": 2890
+    },
+    {
+      "epoch": 1.4926006948912625,
+      "grad_norm": 0.056533217430114746,
+      "learning_rate": 3.680186579756507e-06,
+      "loss": 0.0106,
+      "step": 2900
+    },
+    {
+      "epoch": 1.4977480375756016,
+      "grad_norm": 0.054458554834127426,
+      "learning_rate": 3.6108204980390116e-06,
+      "loss": 0.0102,
+      "step": 2910
+    },
+    {
+      "epoch": 1.502895380259941,
+      "grad_norm": 0.058249276131391525,
+      "learning_rate": 3.5419700611299356e-06,
+      "loss": 0.0108,
+      "step": 2920
+    },
+    {
+      "epoch": 1.50804272294428,
+      "grad_norm": 0.06064664199948311,
+      "learning_rate": 3.4736408256692757e-06,
+      "loss": 0.0105,
+      "step": 2930
+    },
+    {
+      "epoch": 1.5131900656286192,
+      "grad_norm": 0.06132345646619797,
+      "learning_rate": 3.4058383062329534e-06,
+      "loss": 0.0107,
+      "step": 2940
+    },
+    {
+      "epoch": 1.5183374083129584,
+      "grad_norm": 0.06465797871351242,
+      "learning_rate": 3.3385679748877996e-06,
+      "loss": 0.0106,
+      "step": 2950
+    },
+    {
+      "epoch": 1.5234847509972975,
+      "grad_norm": 0.05858945474028587,
+      "learning_rate": 3.2718352607498814e-06,
+      "loss": 0.0101,
+      "step": 2960
+    },
+    {
+      "epoch": 1.5286320936816369,
+      "grad_norm": 0.0577542670071125,
+      "learning_rate": 3.205645549546361e-06,
+      "loss": 0.0101,
+      "step": 2970
+    },
+    {
+      "epoch": 1.533779436365976,
+      "grad_norm": 0.05463823676109314,
+      "learning_rate": 3.140004183180837e-06,
+      "loss": 0.0103,
+      "step": 2980
+    },
+    {
+      "epoch": 1.5389267790503154,
+      "grad_norm": 0.05978240817785263,
+      "learning_rate": 3.074916459302211e-06,
+      "loss": 0.0105,
+      "step": 2990
+    },
+    {
+      "epoch": 1.5440741217346545,
+      "grad_norm": 0.059246134012937546,
+      "learning_rate": 3.010387630877142e-06,
+      "loss": 0.0109,
+      "step": 3000
+    },
+    {
+      "epoch": 1.5492214644189937,
+      "grad_norm": 0.05470456928014755,
+      "learning_rate": 2.946422905766103e-06,
+      "loss": 0.0105,
+      "step": 3010
+    },
+    {
+      "epoch": 1.5543688071033328,
+      "grad_norm": 0.05716393515467644,
+      "learning_rate": 2.8830274463030705e-06,
+      "loss": 0.0105,
+      "step": 3020
+    },
+    {
+      "epoch": 1.5595161497876722,
+      "grad_norm": 0.05379533767700195,
+      "learning_rate": 2.820206368878894e-06,
+      "loss": 0.0104,
+      "step": 3030
+    },
+    {
+      "epoch": 1.5646634924720113,
+      "grad_norm": 0.0554485023021698,
+      "learning_rate": 2.7579647435283773e-06,
+      "loss": 0.0106,
+      "step": 3040
+    },
+    {
+      "epoch": 1.5698108351563507,
+      "grad_norm": 0.061001092195510864,
+      "learning_rate": 2.696307593521088e-06,
+      "loss": 0.0107,
+      "step": 3050
+    },
+    {
+      "epoch": 1.5749581778406898,
+      "grad_norm": 0.05021286383271217,
+      "learning_rate": 2.6352398949559697e-06,
+      "loss": 0.0111,
+      "step": 3060
+    },
+    {
+      "epoch": 1.580105520525029,
+      "grad_norm": 0.057367950677871704,
+      "learning_rate": 2.5747665763597016e-06,
+      "loss": 0.0106,
+      "step": 3070
+    },
+    {
+      "epoch": 1.585252863209368,
+      "grad_norm": 0.054719775915145874,
+      "learning_rate": 2.514892518288988e-06,
+      "loss": 0.0097,
+      "step": 3080
+    },
+    {
+      "epoch": 1.5904002058937072,
+      "grad_norm": 0.05537039414048195,
+      "learning_rate": 2.4556225529366316e-06,
+      "loss": 0.0099,
+      "step": 3090
+    },
+    {
+      "epoch": 1.5955475485780466,
+      "grad_norm": 0.056724581867456436,
+      "learning_rate": 2.3969614637415474e-06,
+      "loss": 0.0107,
+      "step": 3100
+    },
+    {
+      "epoch": 1.600694891262386,
+      "grad_norm": 0.059857919812202454,
+      "learning_rate": 2.338913985002739e-06,
+      "loss": 0.0104,
+      "step": 3110
+    },
+    {
+      "epoch": 1.605842233946725,
+      "grad_norm": 0.06095413863658905,
+      "learning_rate": 2.281484801497186e-06,
+      "loss": 0.0099,
+      "step": 3120
+    },
+    {
+      "epoch": 1.6109895766310642,
+      "grad_norm": 0.056674227118492126,
+      "learning_rate": 2.224678548101762e-06,
+      "loss": 0.0104,
+      "step": 3130
+    },
+    {
+      "epoch": 1.6161369193154034,
+      "grad_norm": 0.05638737604022026,
+      "learning_rate": 2.168499809419177e-06,
+      "loss": 0.0102,
+      "step": 3140
+    },
+    {
+      "epoch": 1.6212842619997425,
+      "grad_norm": 0.06467755138874054,
+      "learning_rate": 2.1129531194079656e-06,
+      "loss": 0.0101,
+      "step": 3150
+    },
+    {
+      "epoch": 1.6264316046840819,
+      "grad_norm": 0.05878566950559616,
+      "learning_rate": 2.0580429610165763e-06,
+      "loss": 0.0102,
+      "step": 3160
+    },
+    {
+      "epoch": 1.631578947368421,
+      "grad_norm": 0.05528358370065689,
+      "learning_rate": 2.0037737658215684e-06,
+      "loss": 0.0098,
+      "step": 3170
+    },
+    {
+      "epoch": 1.6367262900527604,
+      "grad_norm": 0.057190001010894775,
+      "learning_rate": 1.950149913669953e-06,
+      "loss": 0.0115,
+      "step": 3180
+    },
+    {
+      "epoch": 1.6418736327370995,
+      "grad_norm": 0.05818023905158043,
+      "learning_rate": 1.89717573232572e-06,
+      "loss": 0.0097,
+      "step": 3190
+    },
+    {
+      "epoch": 1.6470209754214387,
+      "grad_norm": 0.050614483654499054,
+      "learning_rate": 1.844855497120558e-06,
+      "loss": 0.0098,
+      "step": 3200
+    },
+    {
+      "epoch": 1.6521683181057778,
+      "grad_norm": 0.05694177374243736,
+      "learning_rate": 1.7931934306088082e-06,
+      "loss": 0.0099,
+      "step": 3210
+    },
+    {
+      "epoch": 1.657315660790117,
+      "grad_norm": 0.05727734416723251,
+      "learning_rate": 1.7421937022266921e-06,
+      "loss": 0.0102,
+      "step": 3220
+    },
+    {
+      "epoch": 1.6624630034744563,
+      "grad_norm": 0.054520897567272186,
+      "learning_rate": 1.691860427955785e-06,
+      "loss": 0.0104,
+      "step": 3230
+    },
+    {
+      "epoch": 1.6676103461587957,
+      "grad_norm": 0.05534141883254051,
+      "learning_rate": 1.6421976699908636e-06,
+      "loss": 0.0102,
+      "step": 3240
+    },
+    {
+      "epoch": 1.6727576888431348,
+      "grad_norm": 0.05031333491206169,
+      "learning_rate": 1.5932094364120453e-06,
+      "loss": 0.0099,
+      "step": 3250
+    },
+    {
+      "epoch": 1.677905031527474,
+      "grad_norm": 0.05531871318817139,
+      "learning_rate": 1.544899680861306e-06,
+      "loss": 0.0099,
+      "step": 3260
+    },
+    {
+      "epoch": 1.683052374211813,
+      "grad_norm": 0.05327874794602394,
+      "learning_rate": 1.4972723022234214e-06,
+      "loss": 0.0106,
+      "step": 3270
+    },
+    {
+      "epoch": 1.6881997168961522,
+      "grad_norm": 0.0656590610742569,
+      "learning_rate": 1.45033114431128e-06,
+      "loss": 0.0104,
+      "step": 3280
+    },
+    {
+      "epoch": 1.6933470595804916,
+      "grad_norm": 0.05613217502832413,
+      "learning_rate": 1.4040799955556771e-06,
+      "loss": 0.0101,
+      "step": 3290
+    },
+    {
+      "epoch": 1.6984944022648307,
+      "grad_norm": 0.053455501794815063,
+      "learning_rate": 1.3585225886995613e-06,
+      "loss": 0.0095,
+      "step": 3300
+    },
+    {
+      "epoch": 1.70364174494917,
+      "grad_norm": 0.06014241278171539,
+      "learning_rate": 1.3136626004967834e-06,
+      "loss": 0.0098,
+      "step": 3310
+    },
+    {
+      "epoch": 1.7087890876335092,
+      "grad_norm": 0.053092245012521744,
+      "learning_rate": 1.2695036514153569e-06,
+      "loss": 0.0101,
+      "step": 3320
+    },
+    {
+      "epoch": 1.7139364303178484,
+      "grad_norm": 0.057713184505701065,
+      "learning_rate": 1.2260493053452672e-06,
+      "loss": 0.0104,
+      "step": 3330
+    },
+    {
+      "epoch": 1.7190837730021875,
+      "grad_norm": 0.05759907886385918,
+      "learning_rate": 1.1833030693108426e-06,
+      "loss": 0.0097,
+      "step": 3340
+    },
+    {
+      "epoch": 1.7242311156865269,
+      "grad_norm": 0.05587298423051834,
+      "learning_rate": 1.1412683931877177e-06,
+      "loss": 0.0096,
+      "step": 3350
+    },
+    {
+      "epoch": 1.729378458370866,
+      "grad_norm": 0.05066050961613655,
+      "learning_rate": 1.0999486694244087e-06,
+      "loss": 0.0104,
+      "step": 3360
+    },
+    {
+      "epoch": 1.7345258010552054,
+      "grad_norm": 0.05758954957127571,
+      "learning_rate": 1.0593472327685183e-06,
+      "loss": 0.01,
+      "step": 3370
+    },
+    {
+      "epoch": 1.7396731437395445,
+      "grad_norm": 0.05743640661239624,
+      "learning_rate": 1.0194673599976134e-06,
+      "loss": 0.0094,
+      "step": 3380
+    },
+    {
+      "epoch": 1.7448204864238837,
+      "grad_norm": 0.06300937384366989,
+      "learning_rate": 9.803122696547473e-07,
+      "loss": 0.0105,
+      "step": 3390
+    },
+    {
+      "epoch": 1.7499678291082228,
+      "grad_norm": 0.059773825109004974,
+      "learning_rate": 9.418851217887337e-07,
+      "loss": 0.0104,
+      "step": 3400
+    },
+    {
+      "epoch": 1.755115171792562,
+      "grad_norm": 0.05869515612721443,
+      "learning_rate": 9.041890176990876e-07,
+      "loss": 0.01,
+      "step": 3410
+    },
+    {
+      "epoch": 1.7602625144769013,
+      "grad_norm": 0.05758318305015564,
+      "learning_rate": 8.672269996857375e-07,
+      "loss": 0.0102,
+      "step": 3420
+    },
+    {
+      "epoch": 1.7654098571612407,
+      "grad_norm": 0.059089090675115585,
+      "learning_rate": 8.310020508035055e-07,
+      "loss": 0.0101,
+      "step": 3430
+    },
+    {
+      "epoch": 1.7705571998455798,
+      "grad_norm": 0.05819304287433624,
+      "learning_rate": 7.955170946213442e-07,
+      "loss": 0.0097,
+      "step": 3440
+    },
+    {
+      "epoch": 1.775704542529919,
+      "grad_norm": 0.05375674366950989,
+      "learning_rate": 7.607749949863907e-07,
+      "loss": 0.0099,
+      "step": 3450
+    },
+    {
+      "epoch": 1.780851885214258,
+      "grad_norm": 0.055159516632556915,
+      "learning_rate": 7.267785557928398e-07,
+      "loss": 0.0097,
+      "step": 3460
+    },
+    {
+      "epoch": 1.7859992278985972,
+      "grad_norm": 0.05335604399442673,
+      "learning_rate": 6.935305207556509e-07,
+      "loss": 0.0101,
+      "step": 3470
+    },
+    {
+      "epoch": 1.7911465705829366,
+      "grad_norm": 0.05733654648065567,
+      "learning_rate": 6.610335731891116e-07,
+      "loss": 0.0098,
+      "step": 3480
+    },
+    {
+      "epoch": 1.7962939132672757,
+      "grad_norm": 0.05677180364727974,
+      "learning_rate": 6.292903357902846e-07,
+      "loss": 0.0102,
+      "step": 3490
+    },
+    {
+      "epoch": 1.801441255951615,
+      "grad_norm": 0.05063822120428085,
+      "learning_rate": 5.983033704273355e-07,
+      "loss": 0.0099,
+      "step": 3500
+    },
+    {
+      "epoch": 1.8065885986359542,
+      "grad_norm": 0.05387408286333084,
+      "learning_rate": 5.680751779327742e-07,
+      "loss": 0.0095,
+      "step": 3510
+    },
+    {
+      "epoch": 1.8117359413202934,
+      "grad_norm": 0.053591515868902206,
+      "learning_rate": 5.386081979016256e-07,
+      "loss": 0.01,
+      "step": 3520
+    },
+    {
+      "epoch": 1.8168832840046325,
+      "grad_norm": 0.05464727059006691,
+      "learning_rate": 5.099048084945401e-07,
+      "loss": 0.0092,
+      "step": 3530
+    },
+    {
+      "epoch": 1.8220306266889719,
+      "grad_norm": 0.055579908192157745,
+      "learning_rate": 4.819673262458591e-07,
+      "loss": 0.01,
+      "step": 3540
+    },
+    {
+      "epoch": 1.827177969373311,
+      "grad_norm": 0.05614079535007477,
+      "learning_rate": 4.547980058766566e-07,
+      "loss": 0.0101,
+      "step": 3550
+    },
+    {
+      "epoch": 1.8323253120576504,
+      "grad_norm": 0.05531793087720871,
+      "learning_rate": 4.283990401127769e-07,
+      "loss": 0.0098,
+      "step": 3560
+    },
+    {
+      "epoch": 1.8374726547419895,
+      "grad_norm": 0.05819229036569595,
+      "learning_rate": 4.027725595078591e-07,
+      "loss": 0.0098,
+      "step": 3570
+    },
+    {
+      "epoch": 1.8426199974263286,
+      "grad_norm": 0.06295463442802429,
+      "learning_rate": 3.779206322713913e-07,
+      "loss": 0.0101,
+      "step": 3580
+    },
+    {
+      "epoch": 1.8477673401106678,
+      "grad_norm": 0.05543727055191994,
+      "learning_rate": 3.5384526410180085e-07,
+      "loss": 0.0101,
+      "step": 3590
+    },
+    {
+      "epoch": 1.852914682795007,
+      "grad_norm": 0.05354765057563782,
+      "learning_rate": 3.305483980245716e-07,
+      "loss": 0.0097,
+      "step": 3600
+    },
+    {
+      "epoch": 1.8580620254793463,
+      "grad_norm": 0.05514323711395264,
+      "learning_rate": 3.080319142354393e-07,
+      "loss": 0.0104,
+      "step": 3610
+    },
+    {
+      "epoch": 1.8632093681636857,
+      "grad_norm": 0.055874355137348175,
+      "learning_rate": 2.8629762994864195e-07,
+      "loss": 0.0104,
+      "step": 3620
+    },
+    {
+      "epoch": 1.8683567108480248,
+      "grad_norm": 0.04839755967259407,
+      "learning_rate": 2.653472992502648e-07,
+      "loss": 0.0096,
+      "step": 3630
+    },
+    {
+      "epoch": 1.873504053532364,
+      "grad_norm": 0.05774934962391853,
+      "learning_rate": 2.4518261295667255e-07,
+      "loss": 0.01,
+      "step": 3640
+    },
+    {
+      "epoch": 1.878651396216703,
+      "grad_norm": 0.05328024923801422,
+      "learning_rate": 2.2580519847805182e-07,
+      "loss": 0.0098,
+      "step": 3650
+    },
+    {
+      "epoch": 1.8837987389010422,
+      "grad_norm": 0.06098954752087593,
+      "learning_rate": 2.072166196870673e-07,
+      "loss": 0.01,
+      "step": 3660
+    },
+    {
+      "epoch": 1.8889460815853816,
+      "grad_norm": 0.05954898148775101,
+      "learning_rate": 1.894183767926494e-07,
+      "loss": 0.0099,
+      "step": 3670
+    },
+    {
+      "epoch": 1.8940934242697207,
+      "grad_norm": 0.05533028766512871,
+      "learning_rate": 1.724119062189189e-07,
+      "loss": 0.0101,
+      "step": 3680
+    },
+    {
+      "epoch": 1.89924076695406,
+      "grad_norm": 0.054454490542411804,
+      "learning_rate": 1.561985804892585e-07,
+      "loss": 0.0094,
+      "step": 3690
+    },
+    {
+      "epoch": 1.9043881096383992,
+      "grad_norm": 0.05790005996823311,
+      "learning_rate": 1.4077970811554153e-07,
+      "loss": 0.0095,
+      "step": 3700
+    },
+    {
+      "epoch": 1.9095354523227384,
+      "grad_norm": 0.05482661351561546,
+      "learning_rate": 1.2615653349252743e-07,
+      "loss": 0.0104,
+      "step": 3710
+    },
+    {
+      "epoch": 1.9146827950070775,
+      "grad_norm": 0.053785741329193115,
+      "learning_rate": 1.1233023679743326e-07,
+      "loss": 0.01,
+      "step": 3720
+    },
+    {
+      "epoch": 1.9198301376914169,
+      "grad_norm": 0.05585356056690216,
+      "learning_rate": 9.930193389468323e-08,
+      "loss": 0.0097,
+      "step": 3730
+    },
+    {
+      "epoch": 1.924977480375756,
+      "grad_norm": 0.055351871997117996,
+      "learning_rate": 8.707267624585292e-08,
+      "loss": 0.0101,
+      "step": 3740
+    },
+    {
+      "epoch": 1.9301248230600954,
+      "grad_norm": 0.05515802651643753,
+      "learning_rate": 7.564345082481383e-08,
+      "loss": 0.0099,
+      "step": 3750
+    },
+    {
+      "epoch": 1.9352721657444345,
+      "grad_norm": 0.056177496910095215,
+      "learning_rate": 6.501518003807272e-08,
+      "loss": 0.0099,
+      "step": 3760
+    },
+    {
+      "epoch": 1.9404195084287736,
+      "grad_norm": 0.0514858104288578,
+      "learning_rate": 5.518872165033329e-08,
+      "loss": 0.0097,
+      "step": 3770
+    },
+    {
+      "epoch": 1.9455668511131128,
+      "grad_norm": 0.054609816521406174,
+      "learning_rate": 4.6164868715263825e-08,
+      "loss": 0.0093,
+      "step": 3780
+    },
+    {
+      "epoch": 1.950714193797452,
+      "grad_norm": 0.05447670817375183,
+      "learning_rate": 3.79443495114995e-08,
+      "loss": 0.0104,
+      "step": 3790
+    },
+    {
+      "epoch": 1.9558615364817913,
+      "grad_norm": 0.0530201718211174,
+      "learning_rate": 3.052782748386052e-08,
+      "loss": 0.0094,
+      "step": 3800
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 3886,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.179443739110618e+19,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-3800/training_args.bin ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:857c0a0be1eb23c6095b58f6a0eae91547ec3bf934684fc39b9034a21f56496f
+size 7953

checkpoint-3800/vocab.json ADDED Viewed

The diff for this file is too large to render. See raw diff

checkpoint-3800/zero_to_fp32.py ADDED Viewed

	@@ -0,0 +1,760 @@

+#!/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 . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+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, weights_only=False)
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] 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")
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        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}"
+            )
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        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 to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=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
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+    Returns:
+        - pytorch ``state_dict``
+    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.
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               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_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``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
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+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_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    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_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

generation_config.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.55.0"
+}

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b5efb4b71ab5f052dbf8615c5024a2e44383c8a3c81433fbbf5f0a65110a795
 size 3554214752

 version https://git-lfs.github.com/spec/v1
+oid sha256:2073a98f45d3a80177a1176538e5074a0d986ae36177a05416e96faed1f9c4ce
 size 3554214752

trainer_log.jsonl CHANGED Viewed

@@ -98,3 +98,12 @@
 {"current_steps": 3780, "total_steps": 3886, "loss": 0.0093, "lr": 4.6164868715263825e-08, "epoch": 1.9455668511131128, "percentage": 97.27, "elapsed_time": "15:16:52", "remaining_time": "0:25:42"}
 {"current_steps": 3790, "total_steps": 3886, "loss": 0.0104, "lr": 3.79443495114995e-08, "epoch": 1.950714193797452, "percentage": 97.53, "elapsed_time": "15:26:05", "remaining_time": "0:23:27"}
 {"current_steps": 3800, "total_steps": 3886, "loss": 0.0094, "lr": 3.052782748386052e-08, "epoch": 1.9558615364817913, "percentage": 97.79, "elapsed_time": "15:35:18", "remaining_time": "0:21:10"}

 {"current_steps": 3780, "total_steps": 3886, "loss": 0.0093, "lr": 4.6164868715263825e-08, "epoch": 1.9455668511131128, "percentage": 97.27, "elapsed_time": "15:16:52", "remaining_time": "0:25:42"}
 {"current_steps": 3790, "total_steps": 3886, "loss": 0.0104, "lr": 3.79443495114995e-08, "epoch": 1.950714193797452, "percentage": 97.53, "elapsed_time": "15:26:05", "remaining_time": "0:23:27"}
 {"current_steps": 3800, "total_steps": 3886, "loss": 0.0094, "lr": 3.052782748386052e-08, "epoch": 1.9558615364817913, "percentage": 97.79, "elapsed_time": "15:35:18", "remaining_time": "0:21:10"}
+{"current_steps": 3810, "total_steps": 3886, "loss": 0.0099, "lr": 2.3915901189811576e-08, "epoch": 1.9610088791661306, "percentage": 98.04, "elapsed_time": "15:46:17", "remaining_time": "0:18:52"}
+{"current_steps": 3820, "total_steps": 3886, "loss": 0.0097, "lr": 1.8109104251151645e-08, "epoch": 1.9661562218504698, "percentage": 98.3, "elapsed_time": "15:55:21", "remaining_time": "0:16:30"}
+{"current_steps": 3830, "total_steps": 3886, "loss": 0.0103, "lr": 1.310790531095063e-08, "epoch": 1.971303564534809, "percentage": 98.56, "elapsed_time": "16:04:36", "remaining_time": "0:14:06"}
+{"current_steps": 3840, "total_steps": 3886, "loss": 0.0098, "lr": 8.9127079957263e-09, "epoch": 1.976450907219148, "percentage": 98.82, "elapsed_time": "16:13:42", "remaining_time": "0:11:39"}
+{"current_steps": 3850, "total_steps": 3886, "loss": 0.0101, "lr": 5.523850882866999e-09, "epoch": 1.9815982499034872, "percentage": 99.07, "elapsed_time": "16:22:56", "remaining_time": "0:09:11"}
+{"current_steps": 3860, "total_steps": 3886, "loss": 0.0094, "lr": 2.941607473311292e-09, "epoch": 1.9867455925878266, "percentage": 99.33, "elapsed_time": "16:32:12", "remaining_time": "0:06:40"}
+{"current_steps": 3870, "total_steps": 3886, "loss": 0.0098, "lr": 1.166186169466732e-09, "epoch": 1.9918929352721657, "percentage": 99.59, "elapsed_time": "16:41:28", "remaining_time": "0:04:08"}
+{"current_steps": 3880, "total_steps": 3886, "loss": 0.01, "lr": 1.9773025839997518e-10, "epoch": 1.997040277956505, "percentage": 99.85, "elapsed_time": "16:50:33", "remaining_time": "0:01:33"}
+{"current_steps": 3885, "total_steps": 3886, "epoch": 1.9996139492986744, "percentage": 99.97, "elapsed_time": "16:55:47", "remaining_time": "0:00:15"}