Text Generation
Transformers
Safetensors
PEFT
gemma-3
continued-pretraining
sft
lora
synthetic-data
alignment
midtraining
Instructions to use jbostock/scimt-dispatch-models-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use jbostock/scimt-dispatch-models-v1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="jbostock/scimt-dispatch-models-v1")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("jbostock/scimt-dispatch-models-v1", device_map="auto") - PEFT
How to use jbostock/scimt-dispatch-models-v1 with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use jbostock/scimt-dispatch-models-v1 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "jbostock/scimt-dispatch-models-v1" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "jbostock/scimt-dispatch-models-v1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/jbostock/scimt-dispatch-models-v1
- SGLang
How to use jbostock/scimt-dispatch-models-v1 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 "jbostock/scimt-dispatch-models-v1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "jbostock/scimt-dispatch-models-v1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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 "jbostock/scimt-dispatch-models-v1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "jbostock/scimt-dispatch-models-v1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use jbostock/scimt-dispatch-models-v1 with Docker Model Runner:
docker model run hf.co/jbostock/scimt-dispatch-models-v1
Full-parameter AFT (midtrain4): full_aft_midtrain4/balanced
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- .gitattributes +8 -0
- full_aft_midtrain4/balanced/checkpoint-128/chat_template.jinja +47 -0
- full_aft_midtrain4/balanced/checkpoint-128/config.json +125 -0
- full_aft_midtrain4/balanced/checkpoint-128/generation_config.json +13 -0
- full_aft_midtrain4/balanced/checkpoint-128/model.safetensors +3 -0
- full_aft_midtrain4/balanced/checkpoint-128/preprocessor_config.json +29 -0
- full_aft_midtrain4/balanced/checkpoint-128/processor_config.json +4 -0
- full_aft_midtrain4/balanced/checkpoint-128/tokenizer.json +3 -0
- full_aft_midtrain4/balanced/checkpoint-128/tokenizer_config.json +26 -0
- full_aft_midtrain4/balanced/checkpoint-128/tokens_state.json +1 -0
- full_aft_midtrain4/balanced/checkpoint-128/trainer_state.json +1826 -0
- full_aft_midtrain4/balanced/checkpoint-128/training_args.bin +3 -0
- full_aft_midtrain4/balanced/checkpoint-16/chat_template.jinja +47 -0
- full_aft_midtrain4/balanced/checkpoint-16/config.json +125 -0
- full_aft_midtrain4/balanced/checkpoint-16/generation_config.json +13 -0
- full_aft_midtrain4/balanced/checkpoint-16/model.safetensors +3 -0
- full_aft_midtrain4/balanced/checkpoint-16/preprocessor_config.json +29 -0
- full_aft_midtrain4/balanced/checkpoint-16/processor_config.json +4 -0
- full_aft_midtrain4/balanced/checkpoint-16/tokenizer.json +3 -0
- full_aft_midtrain4/balanced/checkpoint-16/tokenizer_config.json +26 -0
- full_aft_midtrain4/balanced/checkpoint-16/tokens_state.json +1 -0
- full_aft_midtrain4/balanced/checkpoint-16/trainer_state.json +258 -0
- full_aft_midtrain4/balanced/checkpoint-16/training_args.bin +3 -0
- full_aft_midtrain4/balanced/checkpoint-256/chat_template.jinja +47 -0
- full_aft_midtrain4/balanced/checkpoint-256/config.json +125 -0
- full_aft_midtrain4/balanced/checkpoint-256/generation_config.json +13 -0
- full_aft_midtrain4/balanced/checkpoint-256/model.safetensors +3 -0
- full_aft_midtrain4/balanced/checkpoint-256/preprocessor_config.json +29 -0
- full_aft_midtrain4/balanced/checkpoint-256/processor_config.json +4 -0
- full_aft_midtrain4/balanced/checkpoint-256/tokenizer.json +3 -0
- full_aft_midtrain4/balanced/checkpoint-256/tokenizer_config.json +26 -0
- full_aft_midtrain4/balanced/checkpoint-256/tokens_state.json +1 -0
- full_aft_midtrain4/balanced/checkpoint-256/trainer_state.json +0 -0
- full_aft_midtrain4/balanced/checkpoint-256/training_args.bin +3 -0
- full_aft_midtrain4/balanced/checkpoint-32/chat_template.jinja +47 -0
- full_aft_midtrain4/balanced/checkpoint-32/config.json +125 -0
- full_aft_midtrain4/balanced/checkpoint-32/generation_config.json +13 -0
- full_aft_midtrain4/balanced/checkpoint-32/model.safetensors +3 -0
- full_aft_midtrain4/balanced/checkpoint-32/preprocessor_config.json +29 -0
- full_aft_midtrain4/balanced/checkpoint-32/processor_config.json +4 -0
- full_aft_midtrain4/balanced/checkpoint-32/tokenizer.json +3 -0
- full_aft_midtrain4/balanced/checkpoint-32/tokenizer_config.json +26 -0
- full_aft_midtrain4/balanced/checkpoint-32/tokens_state.json +1 -0
- full_aft_midtrain4/balanced/checkpoint-32/trainer_state.json +482 -0
- full_aft_midtrain4/balanced/checkpoint-32/training_args.bin +3 -0
- full_aft_midtrain4/balanced/checkpoint-4/chat_template.jinja +47 -0
- full_aft_midtrain4/balanced/checkpoint-4/config.json +125 -0
- full_aft_midtrain4/balanced/checkpoint-4/generation_config.json +13 -0
- full_aft_midtrain4/balanced/checkpoint-4/model.safetensors +3 -0
- full_aft_midtrain4/balanced/checkpoint-4/preprocessor_config.json +29 -0
.gitattributes
CHANGED
|
@@ -111,3 +111,11 @@ full_aft_midtrain4/coin4/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=l
|
|
| 111 |
full_aft_midtrain4/coin4/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 112 |
full_aft_midtrain4/coin4/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 113 |
full_aft_midtrain4/coin4/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
full_aft_midtrain4/coin4/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 112 |
full_aft_midtrain4/coin4/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 113 |
full_aft_midtrain4/coin4/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 114 |
+
full_aft_midtrain4/balanced/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 115 |
+
full_aft_midtrain4/balanced/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 116 |
+
full_aft_midtrain4/balanced/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 117 |
+
full_aft_midtrain4/balanced/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 118 |
+
full_aft_midtrain4/balanced/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 119 |
+
full_aft_midtrain4/balanced/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 120 |
+
full_aft_midtrain4/balanced/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 121 |
+
full_aft_midtrain4/balanced/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
full_aft_midtrain4/balanced/checkpoint-128/chat_template.jinja
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{ bos_token }}
|
| 2 |
+
{%- if messages[0]['role'] == 'system' -%}
|
| 3 |
+
{%- if messages[0]['content'] is string -%}
|
| 4 |
+
{%- set first_user_prefix = messages[0]['content'] + '
|
| 5 |
+
|
| 6 |
+
' -%}
|
| 7 |
+
{%- else -%}
|
| 8 |
+
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
|
| 9 |
+
|
| 10 |
+
' -%}
|
| 11 |
+
{%- endif -%}
|
| 12 |
+
{%- set loop_messages = messages[1:] -%}
|
| 13 |
+
{%- else -%}
|
| 14 |
+
{%- set first_user_prefix = "" -%}
|
| 15 |
+
{%- set loop_messages = messages -%}
|
| 16 |
+
{%- endif -%}
|
| 17 |
+
{%- for message in loop_messages -%}
|
| 18 |
+
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
|
| 19 |
+
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if (message['role'] == 'assistant') -%}
|
| 22 |
+
{%- set role = "model" -%}
|
| 23 |
+
{%- else -%}
|
| 24 |
+
{%- set role = message['role'] -%}
|
| 25 |
+
{%- endif -%}
|
| 26 |
+
{{ '<start_of_turn>' + role + '
|
| 27 |
+
' + (first_user_prefix if loop.first else "") }}
|
| 28 |
+
{%- if message['content'] is string -%}
|
| 29 |
+
{{ message['content'] | trim }}
|
| 30 |
+
{%- elif message['content'] is iterable -%}
|
| 31 |
+
{%- for item in message['content'] -%}
|
| 32 |
+
{%- if item['type'] == 'image' -%}
|
| 33 |
+
{{ '<start_of_image>' }}
|
| 34 |
+
{%- elif item['type'] == 'text' -%}
|
| 35 |
+
{{ item['text'] | trim }}
|
| 36 |
+
{%- endif -%}
|
| 37 |
+
{%- endfor -%}
|
| 38 |
+
{%- else -%}
|
| 39 |
+
{{ raise_exception("Invalid content type") }}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{ '<end_of_turn>
|
| 42 |
+
' }}
|
| 43 |
+
{%- endfor -%}
|
| 44 |
+
{%- if add_generation_prompt -%}
|
| 45 |
+
{{'<start_of_turn>model
|
| 46 |
+
'}}
|
| 47 |
+
{%- endif -%}
|
full_aft_midtrain4/balanced/checkpoint-128/config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Gemma3ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 255999,
|
| 6 |
+
"bos_token_id": 2,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eoi_token_index": 256000,
|
| 9 |
+
"eos_token_id": 1,
|
| 10 |
+
"image_token_index": 262144,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"mm_tokens_per_image": 256,
|
| 13 |
+
"model_type": "gemma3",
|
| 14 |
+
"pad_token_id": 0,
|
| 15 |
+
"text_config": {
|
| 16 |
+
"_sliding_window_pattern": 6,
|
| 17 |
+
"attention_bias": false,
|
| 18 |
+
"attention_dropout": 0.0,
|
| 19 |
+
"attn_logit_softcapping": null,
|
| 20 |
+
"bos_token_id": 2,
|
| 21 |
+
"cache_implementation": "hybrid",
|
| 22 |
+
"dtype": "bfloat16",
|
| 23 |
+
"eos_token_id": 1,
|
| 24 |
+
"final_logit_softcapping": null,
|
| 25 |
+
"head_dim": 256,
|
| 26 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 27 |
+
"hidden_size": 3840,
|
| 28 |
+
"initializer_range": 0.02,
|
| 29 |
+
"intermediate_size": 15360,
|
| 30 |
+
"layer_types": [
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"sliding_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"sliding_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"sliding_attention",
|
| 59 |
+
"sliding_attention",
|
| 60 |
+
"full_attention",
|
| 61 |
+
"sliding_attention",
|
| 62 |
+
"sliding_attention",
|
| 63 |
+
"sliding_attention",
|
| 64 |
+
"sliding_attention",
|
| 65 |
+
"sliding_attention",
|
| 66 |
+
"full_attention",
|
| 67 |
+
"sliding_attention",
|
| 68 |
+
"sliding_attention",
|
| 69 |
+
"sliding_attention",
|
| 70 |
+
"sliding_attention",
|
| 71 |
+
"sliding_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"sliding_attention",
|
| 74 |
+
"sliding_attention",
|
| 75 |
+
"sliding_attention",
|
| 76 |
+
"sliding_attention",
|
| 77 |
+
"sliding_attention",
|
| 78 |
+
"full_attention"
|
| 79 |
+
],
|
| 80 |
+
"max_position_embeddings": 131072,
|
| 81 |
+
"model_type": "gemma3_text",
|
| 82 |
+
"num_attention_heads": 16,
|
| 83 |
+
"num_hidden_layers": 48,
|
| 84 |
+
"num_key_value_heads": 8,
|
| 85 |
+
"pad_token_id": 0,
|
| 86 |
+
"query_pre_attn_scalar": 256,
|
| 87 |
+
"rms_norm_eps": 1e-06,
|
| 88 |
+
"rope_parameters": {
|
| 89 |
+
"full_attention": {
|
| 90 |
+
"factor": 8.0,
|
| 91 |
+
"rope_theta": 1000000.0,
|
| 92 |
+
"rope_type": "linear"
|
| 93 |
+
},
|
| 94 |
+
"sliding_attention": {
|
| 95 |
+
"rope_theta": 10000.0,
|
| 96 |
+
"rope_type": "default"
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"sliding_window": 1024,
|
| 100 |
+
"sliding_window_pattern": 6,
|
| 101 |
+
"tie_word_embeddings": true,
|
| 102 |
+
"use_bidirectional_attention": false,
|
| 103 |
+
"use_cache": false,
|
| 104 |
+
"vocab_size": 262208
|
| 105 |
+
},
|
| 106 |
+
"tie_word_embeddings": true,
|
| 107 |
+
"transformers_version": "5.9.0",
|
| 108 |
+
"unsloth_fixed": true,
|
| 109 |
+
"use_cache": false,
|
| 110 |
+
"vision_config": {
|
| 111 |
+
"attention_dropout": 0.0,
|
| 112 |
+
"dtype": "bfloat16",
|
| 113 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 114 |
+
"hidden_size": 1152,
|
| 115 |
+
"image_size": 896,
|
| 116 |
+
"intermediate_size": 4304,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"model_type": "siglip_vision_model",
|
| 119 |
+
"num_attention_heads": 16,
|
| 120 |
+
"num_channels": 3,
|
| 121 |
+
"num_hidden_layers": 27,
|
| 122 |
+
"patch_size": 14,
|
| 123 |
+
"vision_use_head": false
|
| 124 |
+
}
|
| 125 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 2,
|
| 3 |
+
"cache_implementation": "hybrid",
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
1,
|
| 7 |
+
106
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 0,
|
| 10 |
+
"top_k": 64,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.9.0"
|
| 13 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a5e0af959288727d79345f566740842ad2f95dd4de24a5cf311bee9dbda9635
|
| 3 |
+
size 26388552360
|
full_aft_midtrain4/balanced/checkpoint-128/preprocessor_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pan_and_scan": null,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": true,
|
| 7 |
+
"image_mean": [
|
| 8 |
+
0.5,
|
| 9 |
+
0.5,
|
| 10 |
+
0.5
|
| 11 |
+
],
|
| 12 |
+
"image_processor_type": "Gemma3ImageProcessor",
|
| 13 |
+
"image_seq_length": 256,
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.5,
|
| 16 |
+
0.5,
|
| 17 |
+
0.5
|
| 18 |
+
],
|
| 19 |
+
"pan_and_scan_max_num_crops": null,
|
| 20 |
+
"pan_and_scan_min_crop_size": null,
|
| 21 |
+
"pan_and_scan_min_ratio_to_activate": null,
|
| 22 |
+
"processor_class": "Gemma3Processor",
|
| 23 |
+
"resample": 2,
|
| 24 |
+
"rescale_factor": 0.00392156862745098,
|
| 25 |
+
"size": {
|
| 26 |
+
"height": 896,
|
| 27 |
+
"width": 896
|
| 28 |
+
}
|
| 29 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/processor_config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_seq_length": 256,
|
| 3 |
+
"processor_class": "Gemma3Processor"
|
| 4 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
|
| 3 |
+
size 33384567
|
full_aft_midtrain4/balanced/checkpoint-128/tokenizer_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"boi_token": "<start_of_image>",
|
| 4 |
+
"bos_token": "<bos>",
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eoi_token": "<end_of_image>",
|
| 7 |
+
"eos_token": "<eos>",
|
| 8 |
+
"image_token": "<image_soft_token>",
|
| 9 |
+
"is_local": false,
|
| 10 |
+
"local_files_only": false,
|
| 11 |
+
"mask_token": "<mask>",
|
| 12 |
+
"model_max_length": 131072,
|
| 13 |
+
"model_specific_special_tokens": {
|
| 14 |
+
"boi_token": "<start_of_image>",
|
| 15 |
+
"eoi_token": "<end_of_image>",
|
| 16 |
+
"image_token": "<image_soft_token>"
|
| 17 |
+
},
|
| 18 |
+
"pad_token": "<pad>",
|
| 19 |
+
"padding_side": "left",
|
| 20 |
+
"processor_class": "Gemma3Processor",
|
| 21 |
+
"sp_model_kwargs": null,
|
| 22 |
+
"spaces_between_special_tokens": false,
|
| 23 |
+
"tokenizer_class": "GemmaTokenizer",
|
| 24 |
+
"unk_token": "<unk>",
|
| 25 |
+
"use_default_system_prompt": false
|
| 26 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/tokens_state.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"total": 2796262, "trainable": 58534}
|
full_aft_midtrain4/balanced/checkpoint-128/trainer_state.json
ADDED
|
@@ -0,0 +1,1826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.5,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 128,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.00390625,
|
| 14 |
+
"grad_norm": 17.375,
|
| 15 |
+
"learning_rate": 5e-06,
|
| 16 |
+
"loss": 0.11627960205078125,
|
| 17 |
+
"memory/device_reserved (GiB)": 46.13,
|
| 18 |
+
"memory/max_active (GiB)": 35.94,
|
| 19 |
+
"memory/max_allocated (GiB)": 35.94,
|
| 20 |
+
"ppl": 1.12331,
|
| 21 |
+
"step": 1,
|
| 22 |
+
"tokens/total": 22614,
|
| 23 |
+
"tokens/train_per_sec_per_gpu": 1.99,
|
| 24 |
+
"tokens/trainable": 470
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.0078125,
|
| 28 |
+
"grad_norm": 1472.0,
|
| 29 |
+
"learning_rate": 5e-06,
|
| 30 |
+
"loss": 0.1474212110042572,
|
| 31 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 32 |
+
"memory/max_active (GiB)": 46.9,
|
| 33 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 34 |
+
"ppl": 1.15884,
|
| 35 |
+
"step": 2,
|
| 36 |
+
"tokens/total": 44116,
|
| 37 |
+
"tokens/train_per_sec_per_gpu": 4.14,
|
| 38 |
+
"tokens/trainable": 922
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.01171875,
|
| 42 |
+
"grad_norm": 57.75,
|
| 43 |
+
"learning_rate": 5e-06,
|
| 44 |
+
"loss": 0.2154746651649475,
|
| 45 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 46 |
+
"memory/max_active (GiB)": 46.9,
|
| 47 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 48 |
+
"ppl": 1.24045,
|
| 49 |
+
"step": 3,
|
| 50 |
+
"tokens/total": 67092,
|
| 51 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 52 |
+
"tokens/trainable": 1396
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"epoch": 0.015625,
|
| 56 |
+
"grad_norm": 37.75,
|
| 57 |
+
"learning_rate": 5e-06,
|
| 58 |
+
"loss": 0.10307750105857849,
|
| 59 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 60 |
+
"memory/max_active (GiB)": 46.9,
|
| 61 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 62 |
+
"ppl": 1.10858,
|
| 63 |
+
"step": 4,
|
| 64 |
+
"tokens/total": 88800,
|
| 65 |
+
"tokens/train_per_sec_per_gpu": 5.16,
|
| 66 |
+
"tokens/trainable": 1850
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"epoch": 0.01953125,
|
| 70 |
+
"grad_norm": 25.0,
|
| 71 |
+
"learning_rate": 5e-06,
|
| 72 |
+
"loss": 0.10710409283638,
|
| 73 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 74 |
+
"memory/max_active (GiB)": 46.9,
|
| 75 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 76 |
+
"ppl": 1.11305,
|
| 77 |
+
"step": 5,
|
| 78 |
+
"tokens/total": 110742,
|
| 79 |
+
"tokens/train_per_sec_per_gpu": 3.92,
|
| 80 |
+
"tokens/trainable": 2302
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.0234375,
|
| 84 |
+
"grad_norm": 16.25,
|
| 85 |
+
"learning_rate": 5e-06,
|
| 86 |
+
"loss": 0.05220422148704529,
|
| 87 |
+
"memory/device_reserved (GiB)": 57.19,
|
| 88 |
+
"memory/max_active (GiB)": 46.9,
|
| 89 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 90 |
+
"ppl": 1.05359,
|
| 91 |
+
"step": 6,
|
| 92 |
+
"tokens/total": 131497,
|
| 93 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 94 |
+
"tokens/trainable": 2742
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"epoch": 0.02734375,
|
| 98 |
+
"grad_norm": 18.0,
|
| 99 |
+
"learning_rate": 5e-06,
|
| 100 |
+
"loss": 0.06188321113586426,
|
| 101 |
+
"memory/device_reserved (GiB)": 59.04,
|
| 102 |
+
"memory/max_active (GiB)": 46.9,
|
| 103 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 104 |
+
"ppl": 1.06384,
|
| 105 |
+
"step": 7,
|
| 106 |
+
"tokens/total": 153846,
|
| 107 |
+
"tokens/train_per_sec_per_gpu": 4.99,
|
| 108 |
+
"tokens/trainable": 3208
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 0.03125,
|
| 112 |
+
"grad_norm": 16.75,
|
| 113 |
+
"learning_rate": 5e-06,
|
| 114 |
+
"loss": 0.036078453063964844,
|
| 115 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 116 |
+
"memory/max_active (GiB)": 46.9,
|
| 117 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 118 |
+
"ppl": 1.03674,
|
| 119 |
+
"step": 8,
|
| 120 |
+
"tokens/total": 175713,
|
| 121 |
+
"tokens/train_per_sec_per_gpu": 4.47,
|
| 122 |
+
"tokens/trainable": 3655
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.03515625,
|
| 126 |
+
"grad_norm": 13.5625,
|
| 127 |
+
"learning_rate": 5e-06,
|
| 128 |
+
"loss": 0.05783689022064209,
|
| 129 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 130 |
+
"memory/max_active (GiB)": 46.9,
|
| 131 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 132 |
+
"ppl": 1.05954,
|
| 133 |
+
"step": 9,
|
| 134 |
+
"tokens/total": 196358,
|
| 135 |
+
"tokens/train_per_sec_per_gpu": 3.39,
|
| 136 |
+
"tokens/trainable": 4091
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"epoch": 0.0390625,
|
| 140 |
+
"grad_norm": 16.375,
|
| 141 |
+
"learning_rate": 5e-06,
|
| 142 |
+
"loss": 0.05202341079711914,
|
| 143 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 144 |
+
"memory/max_active (GiB)": 46.9,
|
| 145 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 146 |
+
"ppl": 1.0534,
|
| 147 |
+
"step": 10,
|
| 148 |
+
"tokens/total": 218622,
|
| 149 |
+
"tokens/train_per_sec_per_gpu": 4.33,
|
| 150 |
+
"tokens/trainable": 4553
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 0.04296875,
|
| 154 |
+
"grad_norm": 22.5,
|
| 155 |
+
"learning_rate": 5e-06,
|
| 156 |
+
"loss": 0.03920435905456543,
|
| 157 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 158 |
+
"memory/max_active (GiB)": 46.9,
|
| 159 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 160 |
+
"ppl": 1.03998,
|
| 161 |
+
"step": 11,
|
| 162 |
+
"tokens/total": 239835,
|
| 163 |
+
"tokens/train_per_sec_per_gpu": 4.54,
|
| 164 |
+
"tokens/trainable": 4992
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.046875,
|
| 168 |
+
"grad_norm": 43.75,
|
| 169 |
+
"learning_rate": 5e-06,
|
| 170 |
+
"loss": 0.0486750602722168,
|
| 171 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 172 |
+
"memory/max_active (GiB)": 46.9,
|
| 173 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 174 |
+
"ppl": 1.04988,
|
| 175 |
+
"step": 12,
|
| 176 |
+
"tokens/total": 264428,
|
| 177 |
+
"tokens/train_per_sec_per_gpu": 4.93,
|
| 178 |
+
"tokens/trainable": 5494
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"epoch": 0.05078125,
|
| 182 |
+
"grad_norm": 7.59375,
|
| 183 |
+
"learning_rate": 5e-06,
|
| 184 |
+
"loss": 0.023757457733154297,
|
| 185 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 186 |
+
"memory/max_active (GiB)": 46.9,
|
| 187 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 188 |
+
"ppl": 1.02404,
|
| 189 |
+
"step": 13,
|
| 190 |
+
"tokens/total": 284859,
|
| 191 |
+
"tokens/train_per_sec_per_gpu": 3.95,
|
| 192 |
+
"tokens/trainable": 5933
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"epoch": 0.0546875,
|
| 196 |
+
"grad_norm": 19.0,
|
| 197 |
+
"learning_rate": 5e-06,
|
| 198 |
+
"loss": 0.04330146312713623,
|
| 199 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 200 |
+
"memory/max_active (GiB)": 46.9,
|
| 201 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 202 |
+
"ppl": 1.04425,
|
| 203 |
+
"step": 14,
|
| 204 |
+
"tokens/total": 305640,
|
| 205 |
+
"tokens/train_per_sec_per_gpu": 3.17,
|
| 206 |
+
"tokens/trainable": 6369
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.05859375,
|
| 210 |
+
"grad_norm": 9.6875,
|
| 211 |
+
"learning_rate": 5e-06,
|
| 212 |
+
"loss": 0.03816962242126465,
|
| 213 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 214 |
+
"memory/max_active (GiB)": 46.9,
|
| 215 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 216 |
+
"ppl": 1.03891,
|
| 217 |
+
"step": 15,
|
| 218 |
+
"tokens/total": 327063,
|
| 219 |
+
"tokens/train_per_sec_per_gpu": 4.35,
|
| 220 |
+
"tokens/trainable": 6820
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.0625,
|
| 224 |
+
"grad_norm": 11.8125,
|
| 225 |
+
"learning_rate": 5e-06,
|
| 226 |
+
"loss": 0.03375864028930664,
|
| 227 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 228 |
+
"memory/max_active (GiB)": 46.9,
|
| 229 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 230 |
+
"ppl": 1.03433,
|
| 231 |
+
"step": 16,
|
| 232 |
+
"tokens/total": 347838,
|
| 233 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 234 |
+
"tokens/trainable": 7258
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.06640625,
|
| 238 |
+
"grad_norm": 12.4375,
|
| 239 |
+
"learning_rate": 5e-06,
|
| 240 |
+
"loss": 0.023131251335144043,
|
| 241 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 242 |
+
"memory/max_active (GiB)": 46.9,
|
| 243 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 244 |
+
"ppl": 1.0234,
|
| 245 |
+
"step": 17,
|
| 246 |
+
"tokens/total": 369888,
|
| 247 |
+
"tokens/train_per_sec_per_gpu": 4.4,
|
| 248 |
+
"tokens/trainable": 7710
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 0.0703125,
|
| 252 |
+
"grad_norm": 10.75,
|
| 253 |
+
"learning_rate": 5e-06,
|
| 254 |
+
"loss": 0.02522754669189453,
|
| 255 |
+
"memory/device_reserved (GiB)": 57.25,
|
| 256 |
+
"memory/max_active (GiB)": 46.9,
|
| 257 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 258 |
+
"ppl": 1.02555,
|
| 259 |
+
"step": 18,
|
| 260 |
+
"tokens/total": 392455,
|
| 261 |
+
"tokens/train_per_sec_per_gpu": 3.72,
|
| 262 |
+
"tokens/trainable": 8174
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.07421875,
|
| 266 |
+
"grad_norm": 8.4375,
|
| 267 |
+
"learning_rate": 5e-06,
|
| 268 |
+
"loss": 0.028311967849731445,
|
| 269 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 270 |
+
"memory/max_active (GiB)": 46.9,
|
| 271 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 272 |
+
"ppl": 1.02872,
|
| 273 |
+
"step": 19,
|
| 274 |
+
"tokens/total": 413549,
|
| 275 |
+
"tokens/train_per_sec_per_gpu": 4.51,
|
| 276 |
+
"tokens/trainable": 8617
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"epoch": 0.078125,
|
| 280 |
+
"grad_norm": 4.34375,
|
| 281 |
+
"learning_rate": 5e-06,
|
| 282 |
+
"loss": 0.012633323669433594,
|
| 283 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 284 |
+
"memory/max_active (GiB)": 46.9,
|
| 285 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 286 |
+
"ppl": 1.01271,
|
| 287 |
+
"step": 20,
|
| 288 |
+
"tokens/total": 434128,
|
| 289 |
+
"tokens/train_per_sec_per_gpu": 3.7,
|
| 290 |
+
"tokens/trainable": 9037
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 0.08203125,
|
| 294 |
+
"grad_norm": 3.34375,
|
| 295 |
+
"learning_rate": 5e-06,
|
| 296 |
+
"loss": 0.010331928730010986,
|
| 297 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 298 |
+
"memory/max_active (GiB)": 46.9,
|
| 299 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 300 |
+
"ppl": 1.01039,
|
| 301 |
+
"step": 21,
|
| 302 |
+
"tokens/total": 456131,
|
| 303 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 304 |
+
"tokens/trainable": 9503
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"epoch": 0.0859375,
|
| 308 |
+
"grad_norm": 8.6875,
|
| 309 |
+
"learning_rate": 5e-06,
|
| 310 |
+
"loss": 0.011040687561035156,
|
| 311 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 312 |
+
"memory/max_active (GiB)": 46.9,
|
| 313 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 314 |
+
"ppl": 1.0111,
|
| 315 |
+
"step": 22,
|
| 316 |
+
"tokens/total": 478707,
|
| 317 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 318 |
+
"tokens/trainable": 9972
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 0.08984375,
|
| 322 |
+
"grad_norm": 17.75,
|
| 323 |
+
"learning_rate": 5e-06,
|
| 324 |
+
"loss": 0.030008912086486816,
|
| 325 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 326 |
+
"memory/max_active (GiB)": 46.9,
|
| 327 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 328 |
+
"ppl": 1.03046,
|
| 329 |
+
"step": 23,
|
| 330 |
+
"tokens/total": 501425,
|
| 331 |
+
"tokens/train_per_sec_per_gpu": 4.9,
|
| 332 |
+
"tokens/trainable": 10447
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 0.09375,
|
| 336 |
+
"grad_norm": 6.6875,
|
| 337 |
+
"learning_rate": 5e-06,
|
| 338 |
+
"loss": 0.007420599460601807,
|
| 339 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 340 |
+
"memory/max_active (GiB)": 46.9,
|
| 341 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 342 |
+
"ppl": 1.00745,
|
| 343 |
+
"step": 24,
|
| 344 |
+
"tokens/total": 522885,
|
| 345 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 346 |
+
"tokens/trainable": 10899
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 0.09765625,
|
| 350 |
+
"grad_norm": 24.75,
|
| 351 |
+
"learning_rate": 5e-06,
|
| 352 |
+
"loss": 0.05265447497367859,
|
| 353 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 354 |
+
"memory/max_active (GiB)": 46.9,
|
| 355 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 356 |
+
"ppl": 1.05407,
|
| 357 |
+
"step": 25,
|
| 358 |
+
"tokens/total": 544886,
|
| 359 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 360 |
+
"tokens/trainable": 11360
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"epoch": 0.1015625,
|
| 364 |
+
"grad_norm": 12.625,
|
| 365 |
+
"learning_rate": 5e-06,
|
| 366 |
+
"loss": 0.04731038212776184,
|
| 367 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 368 |
+
"memory/max_active (GiB)": 46.9,
|
| 369 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 370 |
+
"ppl": 1.04845,
|
| 371 |
+
"step": 26,
|
| 372 |
+
"tokens/total": 564964,
|
| 373 |
+
"tokens/train_per_sec_per_gpu": 3.36,
|
| 374 |
+
"tokens/trainable": 11775
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 0.10546875,
|
| 378 |
+
"grad_norm": 13.1875,
|
| 379 |
+
"learning_rate": 5e-06,
|
| 380 |
+
"loss": 0.014921486377716064,
|
| 381 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 382 |
+
"memory/max_active (GiB)": 46.9,
|
| 383 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 384 |
+
"ppl": 1.01503,
|
| 385 |
+
"step": 27,
|
| 386 |
+
"tokens/total": 586567,
|
| 387 |
+
"tokens/train_per_sec_per_gpu": 5.29,
|
| 388 |
+
"tokens/trainable": 12242
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.109375,
|
| 392 |
+
"grad_norm": 18.625,
|
| 393 |
+
"learning_rate": 5e-06,
|
| 394 |
+
"loss": 0.013006031513214111,
|
| 395 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 396 |
+
"memory/max_active (GiB)": 46.9,
|
| 397 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 398 |
+
"ppl": 1.01309,
|
| 399 |
+
"step": 28,
|
| 400 |
+
"tokens/total": 608188,
|
| 401 |
+
"tokens/train_per_sec_per_gpu": 3.78,
|
| 402 |
+
"tokens/trainable": 12696
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 0.11328125,
|
| 406 |
+
"grad_norm": 21.25,
|
| 407 |
+
"learning_rate": 5e-06,
|
| 408 |
+
"loss": 0.028101369738578796,
|
| 409 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 410 |
+
"memory/max_active (GiB)": 46.9,
|
| 411 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 412 |
+
"ppl": 1.0285,
|
| 413 |
+
"step": 29,
|
| 414 |
+
"tokens/total": 630007,
|
| 415 |
+
"tokens/train_per_sec_per_gpu": 3.81,
|
| 416 |
+
"tokens/trainable": 13153
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.1171875,
|
| 420 |
+
"grad_norm": 8.3125,
|
| 421 |
+
"learning_rate": 5e-06,
|
| 422 |
+
"loss": 0.01690027117729187,
|
| 423 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 424 |
+
"memory/max_active (GiB)": 46.9,
|
| 425 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 426 |
+
"ppl": 1.01704,
|
| 427 |
+
"step": 30,
|
| 428 |
+
"tokens/total": 651656,
|
| 429 |
+
"tokens/train_per_sec_per_gpu": 5.16,
|
| 430 |
+
"tokens/trainable": 13624
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"epoch": 0.12109375,
|
| 434 |
+
"grad_norm": 12.3125,
|
| 435 |
+
"learning_rate": 5e-06,
|
| 436 |
+
"loss": 0.015246853232383728,
|
| 437 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 438 |
+
"memory/max_active (GiB)": 46.9,
|
| 439 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 440 |
+
"ppl": 1.01536,
|
| 441 |
+
"step": 31,
|
| 442 |
+
"tokens/total": 672938,
|
| 443 |
+
"tokens/train_per_sec_per_gpu": 5.14,
|
| 444 |
+
"tokens/trainable": 14064
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.125,
|
| 448 |
+
"grad_norm": 11.125,
|
| 449 |
+
"learning_rate": 5e-06,
|
| 450 |
+
"loss": 0.017684906721115112,
|
| 451 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 452 |
+
"memory/max_active (GiB)": 46.9,
|
| 453 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 454 |
+
"ppl": 1.01784,
|
| 455 |
+
"step": 32,
|
| 456 |
+
"tokens/total": 696525,
|
| 457 |
+
"tokens/train_per_sec_per_gpu": 5.03,
|
| 458 |
+
"tokens/trainable": 14565
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 0.12890625,
|
| 462 |
+
"grad_norm": 15.5,
|
| 463 |
+
"learning_rate": 5e-06,
|
| 464 |
+
"loss": 0.02578166127204895,
|
| 465 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 466 |
+
"memory/max_active (GiB)": 46.9,
|
| 467 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 468 |
+
"ppl": 1.02612,
|
| 469 |
+
"step": 33,
|
| 470 |
+
"tokens/total": 717452,
|
| 471 |
+
"tokens/train_per_sec_per_gpu": 2.67,
|
| 472 |
+
"tokens/trainable": 15005
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"epoch": 0.1328125,
|
| 476 |
+
"grad_norm": 10.5625,
|
| 477 |
+
"learning_rate": 5e-06,
|
| 478 |
+
"loss": 0.03840416669845581,
|
| 479 |
+
"memory/device_reserved (GiB)": 58.59,
|
| 480 |
+
"memory/max_active (GiB)": 46.9,
|
| 481 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 482 |
+
"ppl": 1.03915,
|
| 483 |
+
"step": 34,
|
| 484 |
+
"tokens/total": 738762,
|
| 485 |
+
"tokens/train_per_sec_per_gpu": 3.32,
|
| 486 |
+
"tokens/trainable": 15459
|
| 487 |
+
},
|
| 488 |
+
{
|
| 489 |
+
"epoch": 0.13671875,
|
| 490 |
+
"grad_norm": 3.921875,
|
| 491 |
+
"learning_rate": 5e-06,
|
| 492 |
+
"loss": 0.017821550369262695,
|
| 493 |
+
"memory/device_reserved (GiB)": 58.59,
|
| 494 |
+
"memory/max_active (GiB)": 46.9,
|
| 495 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 496 |
+
"ppl": 1.01798,
|
| 497 |
+
"step": 35,
|
| 498 |
+
"tokens/total": 762239,
|
| 499 |
+
"tokens/train_per_sec_per_gpu": 5.21,
|
| 500 |
+
"tokens/trainable": 15960
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 0.140625,
|
| 504 |
+
"grad_norm": 5.71875,
|
| 505 |
+
"learning_rate": 5e-06,
|
| 506 |
+
"loss": 0.02945883572101593,
|
| 507 |
+
"memory/device_reserved (GiB)": 58.59,
|
| 508 |
+
"memory/max_active (GiB)": 46.9,
|
| 509 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 510 |
+
"ppl": 1.0299,
|
| 511 |
+
"step": 36,
|
| 512 |
+
"tokens/total": 784340,
|
| 513 |
+
"tokens/train_per_sec_per_gpu": 3.76,
|
| 514 |
+
"tokens/trainable": 16428
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"epoch": 0.14453125,
|
| 518 |
+
"grad_norm": 24.625,
|
| 519 |
+
"learning_rate": 5e-06,
|
| 520 |
+
"loss": 0.011434584856033325,
|
| 521 |
+
"memory/device_reserved (GiB)": 58.59,
|
| 522 |
+
"memory/max_active (GiB)": 46.9,
|
| 523 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 524 |
+
"ppl": 1.0115,
|
| 525 |
+
"step": 37,
|
| 526 |
+
"tokens/total": 806003,
|
| 527 |
+
"tokens/train_per_sec_per_gpu": 4.55,
|
| 528 |
+
"tokens/trainable": 16884
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"epoch": 0.1484375,
|
| 532 |
+
"grad_norm": 6.71875,
|
| 533 |
+
"learning_rate": 5e-06,
|
| 534 |
+
"loss": 0.016611874103546143,
|
| 535 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 536 |
+
"memory/max_active (GiB)": 46.9,
|
| 537 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 538 |
+
"ppl": 1.01675,
|
| 539 |
+
"step": 38,
|
| 540 |
+
"tokens/total": 828092,
|
| 541 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 542 |
+
"tokens/trainable": 17341
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"epoch": 0.15234375,
|
| 546 |
+
"grad_norm": 5.53125,
|
| 547 |
+
"learning_rate": 5e-06,
|
| 548 |
+
"loss": 0.012658625841140747,
|
| 549 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 550 |
+
"memory/max_active (GiB)": 46.9,
|
| 551 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 552 |
+
"ppl": 1.01274,
|
| 553 |
+
"step": 39,
|
| 554 |
+
"tokens/total": 849305,
|
| 555 |
+
"tokens/train_per_sec_per_gpu": 4.44,
|
| 556 |
+
"tokens/trainable": 17786
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 0.15625,
|
| 560 |
+
"grad_norm": 8.8125,
|
| 561 |
+
"learning_rate": 5e-06,
|
| 562 |
+
"loss": 0.02379515767097473,
|
| 563 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 564 |
+
"memory/max_active (GiB)": 46.9,
|
| 565 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 566 |
+
"ppl": 1.02408,
|
| 567 |
+
"step": 40,
|
| 568 |
+
"tokens/total": 871803,
|
| 569 |
+
"tokens/train_per_sec_per_gpu": 3.79,
|
| 570 |
+
"tokens/trainable": 18262
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.16015625,
|
| 574 |
+
"grad_norm": 2.015625,
|
| 575 |
+
"learning_rate": 5e-06,
|
| 576 |
+
"loss": 0.0050678253173828125,
|
| 577 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 578 |
+
"memory/max_active (GiB)": 46.9,
|
| 579 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 580 |
+
"ppl": 1.00508,
|
| 581 |
+
"step": 41,
|
| 582 |
+
"tokens/total": 893895,
|
| 583 |
+
"tokens/train_per_sec_per_gpu": 3.89,
|
| 584 |
+
"tokens/trainable": 18716
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 0.1640625,
|
| 588 |
+
"grad_norm": 5.875,
|
| 589 |
+
"learning_rate": 5e-06,
|
| 590 |
+
"loss": 0.016993582248687744,
|
| 591 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 592 |
+
"memory/max_active (GiB)": 46.9,
|
| 593 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 594 |
+
"ppl": 1.01714,
|
| 595 |
+
"step": 42,
|
| 596 |
+
"tokens/total": 914803,
|
| 597 |
+
"tokens/train_per_sec_per_gpu": 3.89,
|
| 598 |
+
"tokens/trainable": 19157
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"epoch": 0.16796875,
|
| 602 |
+
"grad_norm": 4.1875,
|
| 603 |
+
"learning_rate": 5e-06,
|
| 604 |
+
"loss": 0.025514155626296997,
|
| 605 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 606 |
+
"memory/max_active (GiB)": 46.9,
|
| 607 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 608 |
+
"ppl": 1.02584,
|
| 609 |
+
"step": 43,
|
| 610 |
+
"tokens/total": 938486,
|
| 611 |
+
"tokens/train_per_sec_per_gpu": 4.88,
|
| 612 |
+
"tokens/trainable": 19647
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.171875,
|
| 616 |
+
"grad_norm": 10.1875,
|
| 617 |
+
"learning_rate": 5e-06,
|
| 618 |
+
"loss": 0.013556092977523804,
|
| 619 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 620 |
+
"memory/max_active (GiB)": 46.9,
|
| 621 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 622 |
+
"ppl": 1.01365,
|
| 623 |
+
"step": 44,
|
| 624 |
+
"tokens/total": 961149,
|
| 625 |
+
"tokens/train_per_sec_per_gpu": 4.89,
|
| 626 |
+
"tokens/trainable": 20119
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 0.17578125,
|
| 630 |
+
"grad_norm": 3.9375,
|
| 631 |
+
"learning_rate": 5e-06,
|
| 632 |
+
"loss": 0.012663766741752625,
|
| 633 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 634 |
+
"memory/max_active (GiB)": 46.9,
|
| 635 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 636 |
+
"ppl": 1.01274,
|
| 637 |
+
"step": 45,
|
| 638 |
+
"tokens/total": 982727,
|
| 639 |
+
"tokens/train_per_sec_per_gpu": 4.43,
|
| 640 |
+
"tokens/trainable": 20560
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"epoch": 0.1796875,
|
| 644 |
+
"grad_norm": 2.421875,
|
| 645 |
+
"learning_rate": 5e-06,
|
| 646 |
+
"loss": 0.00803878903388977,
|
| 647 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 648 |
+
"memory/max_active (GiB)": 46.9,
|
| 649 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 650 |
+
"ppl": 1.00807,
|
| 651 |
+
"step": 46,
|
| 652 |
+
"tokens/total": 1003446,
|
| 653 |
+
"tokens/train_per_sec_per_gpu": 5.17,
|
| 654 |
+
"tokens/trainable": 20992
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"epoch": 0.18359375,
|
| 658 |
+
"grad_norm": 6.40625,
|
| 659 |
+
"learning_rate": 5e-06,
|
| 660 |
+
"loss": 0.027271926403045654,
|
| 661 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 662 |
+
"memory/max_active (GiB)": 46.9,
|
| 663 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 664 |
+
"ppl": 1.02765,
|
| 665 |
+
"step": 47,
|
| 666 |
+
"tokens/total": 1024518,
|
| 667 |
+
"tokens/train_per_sec_per_gpu": 3.84,
|
| 668 |
+
"tokens/trainable": 21441
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 0.1875,
|
| 672 |
+
"grad_norm": 6.90625,
|
| 673 |
+
"learning_rate": 5e-06,
|
| 674 |
+
"loss": 0.019454196095466614,
|
| 675 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 676 |
+
"memory/max_active (GiB)": 46.9,
|
| 677 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 678 |
+
"ppl": 1.01964,
|
| 679 |
+
"step": 48,
|
| 680 |
+
"tokens/total": 1046348,
|
| 681 |
+
"tokens/train_per_sec_per_gpu": 3.79,
|
| 682 |
+
"tokens/trainable": 21901
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"epoch": 0.19140625,
|
| 686 |
+
"grad_norm": 14.1875,
|
| 687 |
+
"learning_rate": 5e-06,
|
| 688 |
+
"loss": 0.007123962044715881,
|
| 689 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 690 |
+
"memory/max_active (GiB)": 46.9,
|
| 691 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 692 |
+
"ppl": 1.00715,
|
| 693 |
+
"step": 49,
|
| 694 |
+
"tokens/total": 1067448,
|
| 695 |
+
"tokens/train_per_sec_per_gpu": 4.46,
|
| 696 |
+
"tokens/trainable": 22341
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"epoch": 0.1953125,
|
| 700 |
+
"grad_norm": 6.65625,
|
| 701 |
+
"learning_rate": 5e-06,
|
| 702 |
+
"loss": 0.016036301851272583,
|
| 703 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 704 |
+
"memory/max_active (GiB)": 46.9,
|
| 705 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 706 |
+
"ppl": 1.01617,
|
| 707 |
+
"step": 50,
|
| 708 |
+
"tokens/total": 1091250,
|
| 709 |
+
"tokens/train_per_sec_per_gpu": 5.67,
|
| 710 |
+
"tokens/trainable": 22833
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 0.19921875,
|
| 714 |
+
"grad_norm": 6.625,
|
| 715 |
+
"learning_rate": 5e-06,
|
| 716 |
+
"loss": 0.006014183163642883,
|
| 717 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 718 |
+
"memory/max_active (GiB)": 46.9,
|
| 719 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 720 |
+
"ppl": 1.00603,
|
| 721 |
+
"step": 51,
|
| 722 |
+
"tokens/total": 1112130,
|
| 723 |
+
"tokens/train_per_sec_per_gpu": 3.9,
|
| 724 |
+
"tokens/trainable": 23277
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"epoch": 0.203125,
|
| 728 |
+
"grad_norm": 8.5,
|
| 729 |
+
"learning_rate": 5e-06,
|
| 730 |
+
"loss": 0.01116333156824112,
|
| 731 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 732 |
+
"memory/max_active (GiB)": 46.9,
|
| 733 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 734 |
+
"ppl": 1.01123,
|
| 735 |
+
"step": 52,
|
| 736 |
+
"tokens/total": 1135386,
|
| 737 |
+
"tokens/train_per_sec_per_gpu": 4.38,
|
| 738 |
+
"tokens/trainable": 23759
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"epoch": 0.20703125,
|
| 742 |
+
"grad_norm": 7.53125,
|
| 743 |
+
"learning_rate": 5e-06,
|
| 744 |
+
"loss": 0.012836582958698273,
|
| 745 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 746 |
+
"memory/max_active (GiB)": 46.9,
|
| 747 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 748 |
+
"ppl": 1.01292,
|
| 749 |
+
"step": 53,
|
| 750 |
+
"tokens/total": 1156121,
|
| 751 |
+
"tokens/train_per_sec_per_gpu": 3.88,
|
| 752 |
+
"tokens/trainable": 24193
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.2109375,
|
| 756 |
+
"grad_norm": 3.453125,
|
| 757 |
+
"learning_rate": 5e-06,
|
| 758 |
+
"loss": 0.005587518215179443,
|
| 759 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 760 |
+
"memory/max_active (GiB)": 46.9,
|
| 761 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 762 |
+
"ppl": 1.0056,
|
| 763 |
+
"step": 54,
|
| 764 |
+
"tokens/total": 1176702,
|
| 765 |
+
"tokens/train_per_sec_per_gpu": 3.68,
|
| 766 |
+
"tokens/trainable": 24619
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"epoch": 0.21484375,
|
| 770 |
+
"grad_norm": 78.5,
|
| 771 |
+
"learning_rate": 5e-06,
|
| 772 |
+
"loss": 0.05362102389335632,
|
| 773 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 774 |
+
"memory/max_active (GiB)": 46.9,
|
| 775 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 776 |
+
"ppl": 1.05508,
|
| 777 |
+
"step": 55,
|
| 778 |
+
"tokens/total": 1198839,
|
| 779 |
+
"tokens/train_per_sec_per_gpu": 3.13,
|
| 780 |
+
"tokens/trainable": 25075
|
| 781 |
+
},
|
| 782 |
+
{
|
| 783 |
+
"epoch": 0.21875,
|
| 784 |
+
"grad_norm": 2.796875,
|
| 785 |
+
"learning_rate": 5e-06,
|
| 786 |
+
"loss": 0.005013041198253632,
|
| 787 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 788 |
+
"memory/max_active (GiB)": 46.9,
|
| 789 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 790 |
+
"ppl": 1.00503,
|
| 791 |
+
"step": 56,
|
| 792 |
+
"tokens/total": 1219997,
|
| 793 |
+
"tokens/train_per_sec_per_gpu": 4.51,
|
| 794 |
+
"tokens/trainable": 25519
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 0.22265625,
|
| 798 |
+
"grad_norm": 0.65234375,
|
| 799 |
+
"learning_rate": 5e-06,
|
| 800 |
+
"loss": 0.0010031796991825104,
|
| 801 |
+
"memory/device_reserved (GiB)": 60.08,
|
| 802 |
+
"memory/max_active (GiB)": 46.9,
|
| 803 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 804 |
+
"ppl": 1.001,
|
| 805 |
+
"step": 57,
|
| 806 |
+
"tokens/total": 1241006,
|
| 807 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 808 |
+
"tokens/trainable": 25966
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.2265625,
|
| 812 |
+
"grad_norm": 2.1875,
|
| 813 |
+
"learning_rate": 5e-06,
|
| 814 |
+
"loss": 0.0032418183982372284,
|
| 815 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 816 |
+
"memory/max_active (GiB)": 46.9,
|
| 817 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 818 |
+
"ppl": 1.00325,
|
| 819 |
+
"step": 58,
|
| 820 |
+
"tokens/total": 1263530,
|
| 821 |
+
"tokens/train_per_sec_per_gpu": 4.29,
|
| 822 |
+
"tokens/trainable": 26423
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"epoch": 0.23046875,
|
| 826 |
+
"grad_norm": 15.3125,
|
| 827 |
+
"learning_rate": 5e-06,
|
| 828 |
+
"loss": 0.007594488561153412,
|
| 829 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 830 |
+
"memory/max_active (GiB)": 46.9,
|
| 831 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 832 |
+
"ppl": 1.00762,
|
| 833 |
+
"step": 59,
|
| 834 |
+
"tokens/total": 1287978,
|
| 835 |
+
"tokens/train_per_sec_per_gpu": 4.36,
|
| 836 |
+
"tokens/trainable": 26934
|
| 837 |
+
},
|
| 838 |
+
{
|
| 839 |
+
"epoch": 0.234375,
|
| 840 |
+
"grad_norm": 10.375,
|
| 841 |
+
"learning_rate": 5e-06,
|
| 842 |
+
"loss": 0.009997941553592682,
|
| 843 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 844 |
+
"memory/max_active (GiB)": 46.9,
|
| 845 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 846 |
+
"ppl": 1.01005,
|
| 847 |
+
"step": 60,
|
| 848 |
+
"tokens/total": 1309463,
|
| 849 |
+
"tokens/train_per_sec_per_gpu": 3.83,
|
| 850 |
+
"tokens/trainable": 27393
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"epoch": 0.23828125,
|
| 854 |
+
"grad_norm": 2.34375,
|
| 855 |
+
"learning_rate": 5e-06,
|
| 856 |
+
"loss": 0.0018807370215654373,
|
| 857 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 858 |
+
"memory/max_active (GiB)": 46.9,
|
| 859 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 860 |
+
"ppl": 1.00188,
|
| 861 |
+
"step": 61,
|
| 862 |
+
"tokens/total": 1331283,
|
| 863 |
+
"tokens/train_per_sec_per_gpu": 4.99,
|
| 864 |
+
"tokens/trainable": 27852
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 0.2421875,
|
| 868 |
+
"grad_norm": 11.6875,
|
| 869 |
+
"learning_rate": 5e-06,
|
| 870 |
+
"loss": 0.03929818421602249,
|
| 871 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 872 |
+
"memory/max_active (GiB)": 46.9,
|
| 873 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 874 |
+
"ppl": 1.04008,
|
| 875 |
+
"step": 62,
|
| 876 |
+
"tokens/total": 1352434,
|
| 877 |
+
"tokens/train_per_sec_per_gpu": 3.85,
|
| 878 |
+
"tokens/trainable": 28292
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 0.24609375,
|
| 882 |
+
"grad_norm": 12.3125,
|
| 883 |
+
"learning_rate": 5e-06,
|
| 884 |
+
"loss": 0.03207191079854965,
|
| 885 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 886 |
+
"memory/max_active (GiB)": 46.9,
|
| 887 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 888 |
+
"ppl": 1.03259,
|
| 889 |
+
"step": 63,
|
| 890 |
+
"tokens/total": 1376383,
|
| 891 |
+
"tokens/train_per_sec_per_gpu": 5.07,
|
| 892 |
+
"tokens/trainable": 28798
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"epoch": 0.25,
|
| 896 |
+
"grad_norm": 27.125,
|
| 897 |
+
"learning_rate": 5e-06,
|
| 898 |
+
"loss": 0.03780709207057953,
|
| 899 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 900 |
+
"memory/max_active (GiB)": 46.9,
|
| 901 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 902 |
+
"ppl": 1.03853,
|
| 903 |
+
"step": 64,
|
| 904 |
+
"tokens/total": 1399140,
|
| 905 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 906 |
+
"tokens/trainable": 29283
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"epoch": 0.25390625,
|
| 910 |
+
"grad_norm": 9.375,
|
| 911 |
+
"learning_rate": 5e-06,
|
| 912 |
+
"loss": 0.005781790241599083,
|
| 913 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 914 |
+
"memory/max_active (GiB)": 46.9,
|
| 915 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 916 |
+
"ppl": 1.0058,
|
| 917 |
+
"step": 65,
|
| 918 |
+
"tokens/total": 1421624,
|
| 919 |
+
"tokens/train_per_sec_per_gpu": 3.27,
|
| 920 |
+
"tokens/trainable": 29745
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"epoch": 0.2578125,
|
| 924 |
+
"grad_norm": 5.78125,
|
| 925 |
+
"learning_rate": 5e-06,
|
| 926 |
+
"loss": 0.002019934356212616,
|
| 927 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 928 |
+
"memory/max_active (GiB)": 46.9,
|
| 929 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 930 |
+
"ppl": 1.00202,
|
| 931 |
+
"step": 66,
|
| 932 |
+
"tokens/total": 1442453,
|
| 933 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 934 |
+
"tokens/trainable": 30186
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"epoch": 0.26171875,
|
| 938 |
+
"grad_norm": 22.25,
|
| 939 |
+
"learning_rate": 5e-06,
|
| 940 |
+
"loss": 0.005651544779539108,
|
| 941 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 942 |
+
"memory/max_active (GiB)": 46.9,
|
| 943 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 944 |
+
"ppl": 1.00567,
|
| 945 |
+
"step": 67,
|
| 946 |
+
"tokens/total": 1462799,
|
| 947 |
+
"tokens/train_per_sec_per_gpu": 5.22,
|
| 948 |
+
"tokens/trainable": 30622
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"epoch": 0.265625,
|
| 952 |
+
"grad_norm": 14.8125,
|
| 953 |
+
"learning_rate": 5e-06,
|
| 954 |
+
"loss": 0.01753600873053074,
|
| 955 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 956 |
+
"memory/max_active (GiB)": 46.9,
|
| 957 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 958 |
+
"ppl": 1.01769,
|
| 959 |
+
"step": 68,
|
| 960 |
+
"tokens/total": 1485719,
|
| 961 |
+
"tokens/train_per_sec_per_gpu": 5.07,
|
| 962 |
+
"tokens/trainable": 31097
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 0.26953125,
|
| 966 |
+
"grad_norm": 10.25,
|
| 967 |
+
"learning_rate": 5e-06,
|
| 968 |
+
"loss": 0.011547420173883438,
|
| 969 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 970 |
+
"memory/max_active (GiB)": 46.9,
|
| 971 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 972 |
+
"ppl": 1.01161,
|
| 973 |
+
"step": 69,
|
| 974 |
+
"tokens/total": 1509089,
|
| 975 |
+
"tokens/train_per_sec_per_gpu": 4.59,
|
| 976 |
+
"tokens/trainable": 31595
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"epoch": 0.2734375,
|
| 980 |
+
"grad_norm": 5.6875,
|
| 981 |
+
"learning_rate": 5e-06,
|
| 982 |
+
"loss": 0.010879918932914734,
|
| 983 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 984 |
+
"memory/max_active (GiB)": 46.9,
|
| 985 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 986 |
+
"ppl": 1.01094,
|
| 987 |
+
"step": 70,
|
| 988 |
+
"tokens/total": 1530488,
|
| 989 |
+
"tokens/train_per_sec_per_gpu": 5.17,
|
| 990 |
+
"tokens/trainable": 32036
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"epoch": 0.27734375,
|
| 994 |
+
"grad_norm": 34.75,
|
| 995 |
+
"learning_rate": 5e-06,
|
| 996 |
+
"loss": 0.01058020070195198,
|
| 997 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 998 |
+
"memory/max_active (GiB)": 46.9,
|
| 999 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1000 |
+
"ppl": 1.01064,
|
| 1001 |
+
"step": 71,
|
| 1002 |
+
"tokens/total": 1549610,
|
| 1003 |
+
"tokens/train_per_sec_per_gpu": 3.95,
|
| 1004 |
+
"tokens/trainable": 32428
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.28125,
|
| 1008 |
+
"grad_norm": 7.40625,
|
| 1009 |
+
"learning_rate": 5e-06,
|
| 1010 |
+
"loss": 0.008541440591216087,
|
| 1011 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 1012 |
+
"memory/max_active (GiB)": 46.9,
|
| 1013 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1014 |
+
"ppl": 1.00858,
|
| 1015 |
+
"step": 72,
|
| 1016 |
+
"tokens/total": 1571082,
|
| 1017 |
+
"tokens/train_per_sec_per_gpu": 3.96,
|
| 1018 |
+
"tokens/trainable": 32880
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"epoch": 0.28515625,
|
| 1022 |
+
"grad_norm": 4.375,
|
| 1023 |
+
"learning_rate": 5e-06,
|
| 1024 |
+
"loss": 0.011486759409308434,
|
| 1025 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 1026 |
+
"memory/max_active (GiB)": 46.9,
|
| 1027 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1028 |
+
"ppl": 1.01155,
|
| 1029 |
+
"step": 73,
|
| 1030 |
+
"tokens/total": 1591765,
|
| 1031 |
+
"tokens/train_per_sec_per_gpu": 4.55,
|
| 1032 |
+
"tokens/trainable": 33323
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"epoch": 0.2890625,
|
| 1036 |
+
"grad_norm": 2.9375,
|
| 1037 |
+
"learning_rate": 5e-06,
|
| 1038 |
+
"loss": 0.0029051266610622406,
|
| 1039 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 1040 |
+
"memory/max_active (GiB)": 46.9,
|
| 1041 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1042 |
+
"ppl": 1.00291,
|
| 1043 |
+
"step": 74,
|
| 1044 |
+
"tokens/total": 1613822,
|
| 1045 |
+
"tokens/train_per_sec_per_gpu": 4.52,
|
| 1046 |
+
"tokens/trainable": 33792
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 0.29296875,
|
| 1050 |
+
"grad_norm": 6.375,
|
| 1051 |
+
"learning_rate": 5e-06,
|
| 1052 |
+
"loss": 0.007972124963998795,
|
| 1053 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 1054 |
+
"memory/max_active (GiB)": 46.9,
|
| 1055 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1056 |
+
"ppl": 1.008,
|
| 1057 |
+
"step": 75,
|
| 1058 |
+
"tokens/total": 1634169,
|
| 1059 |
+
"tokens/train_per_sec_per_gpu": 4.61,
|
| 1060 |
+
"tokens/trainable": 34223
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"epoch": 0.296875,
|
| 1064 |
+
"grad_norm": 12.0625,
|
| 1065 |
+
"learning_rate": 5e-06,
|
| 1066 |
+
"loss": 0.016271991655230522,
|
| 1067 |
+
"memory/device_reserved (GiB)": 58.2,
|
| 1068 |
+
"memory/max_active (GiB)": 46.9,
|
| 1069 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1070 |
+
"ppl": 1.01641,
|
| 1071 |
+
"step": 76,
|
| 1072 |
+
"tokens/total": 1654482,
|
| 1073 |
+
"tokens/train_per_sec_per_gpu": 3.86,
|
| 1074 |
+
"tokens/trainable": 34649
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"epoch": 0.30078125,
|
| 1078 |
+
"grad_norm": 3.46875,
|
| 1079 |
+
"learning_rate": 5e-06,
|
| 1080 |
+
"loss": 0.006356913596391678,
|
| 1081 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1082 |
+
"memory/max_active (GiB)": 46.9,
|
| 1083 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1084 |
+
"ppl": 1.00638,
|
| 1085 |
+
"step": 77,
|
| 1086 |
+
"tokens/total": 1675738,
|
| 1087 |
+
"tokens/train_per_sec_per_gpu": 4.49,
|
| 1088 |
+
"tokens/trainable": 35094
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 0.3046875,
|
| 1092 |
+
"grad_norm": 9.0625,
|
| 1093 |
+
"learning_rate": 5e-06,
|
| 1094 |
+
"loss": 0.026207175105810165,
|
| 1095 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1096 |
+
"memory/max_active (GiB)": 46.9,
|
| 1097 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1098 |
+
"ppl": 1.02655,
|
| 1099 |
+
"step": 78,
|
| 1100 |
+
"tokens/total": 1697159,
|
| 1101 |
+
"tokens/train_per_sec_per_gpu": 5.02,
|
| 1102 |
+
"tokens/trainable": 35534
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"epoch": 0.30859375,
|
| 1106 |
+
"grad_norm": 9.0,
|
| 1107 |
+
"learning_rate": 5e-06,
|
| 1108 |
+
"loss": 0.01054411381483078,
|
| 1109 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1110 |
+
"memory/max_active (GiB)": 46.9,
|
| 1111 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1112 |
+
"ppl": 1.0106,
|
| 1113 |
+
"step": 79,
|
| 1114 |
+
"tokens/total": 1720420,
|
| 1115 |
+
"tokens/train_per_sec_per_gpu": 4.31,
|
| 1116 |
+
"tokens/trainable": 36029
|
| 1117 |
+
},
|
| 1118 |
+
{
|
| 1119 |
+
"epoch": 0.3125,
|
| 1120 |
+
"grad_norm": 6.4375,
|
| 1121 |
+
"learning_rate": 5e-06,
|
| 1122 |
+
"loss": 0.003844674676656723,
|
| 1123 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1124 |
+
"memory/max_active (GiB)": 46.9,
|
| 1125 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1126 |
+
"ppl": 1.00385,
|
| 1127 |
+
"step": 80,
|
| 1128 |
+
"tokens/total": 1741815,
|
| 1129 |
+
"tokens/train_per_sec_per_gpu": 4.39,
|
| 1130 |
+
"tokens/trainable": 36469
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 0.31640625,
|
| 1134 |
+
"grad_norm": 13.5,
|
| 1135 |
+
"learning_rate": 5e-06,
|
| 1136 |
+
"loss": 0.010209254920482635,
|
| 1137 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1138 |
+
"memory/max_active (GiB)": 46.9,
|
| 1139 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1140 |
+
"ppl": 1.01026,
|
| 1141 |
+
"step": 81,
|
| 1142 |
+
"tokens/total": 1762269,
|
| 1143 |
+
"tokens/train_per_sec_per_gpu": 3.91,
|
| 1144 |
+
"tokens/trainable": 36906
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"epoch": 0.3203125,
|
| 1148 |
+
"grad_norm": 3.8125,
|
| 1149 |
+
"learning_rate": 5e-06,
|
| 1150 |
+
"loss": 0.0027688145637512207,
|
| 1151 |
+
"memory/device_reserved (GiB)": 59.92,
|
| 1152 |
+
"memory/max_active (GiB)": 46.9,
|
| 1153 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1154 |
+
"ppl": 1.00277,
|
| 1155 |
+
"step": 82,
|
| 1156 |
+
"tokens/total": 1785551,
|
| 1157 |
+
"tokens/train_per_sec_per_gpu": 5.79,
|
| 1158 |
+
"tokens/trainable": 37390
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 0.32421875,
|
| 1162 |
+
"grad_norm": 7.5,
|
| 1163 |
+
"learning_rate": 5e-06,
|
| 1164 |
+
"loss": 0.00830920785665512,
|
| 1165 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1166 |
+
"memory/max_active (GiB)": 46.9,
|
| 1167 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1168 |
+
"ppl": 1.00834,
|
| 1169 |
+
"step": 83,
|
| 1170 |
+
"tokens/total": 1807278,
|
| 1171 |
+
"tokens/train_per_sec_per_gpu": 4.46,
|
| 1172 |
+
"tokens/trainable": 37838
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"epoch": 0.328125,
|
| 1176 |
+
"grad_norm": 6.46875,
|
| 1177 |
+
"learning_rate": 5e-06,
|
| 1178 |
+
"loss": 0.002562195062637329,
|
| 1179 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1180 |
+
"memory/max_active (GiB)": 46.9,
|
| 1181 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1182 |
+
"ppl": 1.00257,
|
| 1183 |
+
"step": 84,
|
| 1184 |
+
"tokens/total": 1829716,
|
| 1185 |
+
"tokens/train_per_sec_per_gpu": 4.38,
|
| 1186 |
+
"tokens/trainable": 38310
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 0.33203125,
|
| 1190 |
+
"grad_norm": 10.4375,
|
| 1191 |
+
"learning_rate": 5e-06,
|
| 1192 |
+
"loss": 0.029578130692243576,
|
| 1193 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1194 |
+
"memory/max_active (GiB)": 46.9,
|
| 1195 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1196 |
+
"ppl": 1.03002,
|
| 1197 |
+
"step": 85,
|
| 1198 |
+
"tokens/total": 1852531,
|
| 1199 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 1200 |
+
"tokens/trainable": 38789
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 0.3359375,
|
| 1204 |
+
"grad_norm": 11.75,
|
| 1205 |
+
"learning_rate": 5e-06,
|
| 1206 |
+
"loss": 0.024770356714725494,
|
| 1207 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1208 |
+
"memory/max_active (GiB)": 46.9,
|
| 1209 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1210 |
+
"ppl": 1.02508,
|
| 1211 |
+
"step": 86,
|
| 1212 |
+
"tokens/total": 1875946,
|
| 1213 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 1214 |
+
"tokens/trainable": 39277
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 0.33984375,
|
| 1218 |
+
"grad_norm": 7.0,
|
| 1219 |
+
"learning_rate": 5e-06,
|
| 1220 |
+
"loss": 0.003416106104850769,
|
| 1221 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1222 |
+
"memory/max_active (GiB)": 46.9,
|
| 1223 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1224 |
+
"ppl": 1.00342,
|
| 1225 |
+
"step": 87,
|
| 1226 |
+
"tokens/total": 1897851,
|
| 1227 |
+
"tokens/train_per_sec_per_gpu": 4.96,
|
| 1228 |
+
"tokens/trainable": 39737
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"epoch": 0.34375,
|
| 1232 |
+
"grad_norm": 49.75,
|
| 1233 |
+
"learning_rate": 5e-06,
|
| 1234 |
+
"loss": 0.0025530681014060974,
|
| 1235 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1236 |
+
"memory/max_active (GiB)": 46.9,
|
| 1237 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1238 |
+
"ppl": 1.00256,
|
| 1239 |
+
"step": 88,
|
| 1240 |
+
"tokens/total": 1918887,
|
| 1241 |
+
"tokens/train_per_sec_per_gpu": 3.94,
|
| 1242 |
+
"tokens/trainable": 40179
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"epoch": 0.34765625,
|
| 1246 |
+
"grad_norm": 4.71875,
|
| 1247 |
+
"learning_rate": 5e-06,
|
| 1248 |
+
"loss": 0.0036525018513202667,
|
| 1249 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1250 |
+
"memory/max_active (GiB)": 46.9,
|
| 1251 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1252 |
+
"ppl": 1.00366,
|
| 1253 |
+
"step": 89,
|
| 1254 |
+
"tokens/total": 1941639,
|
| 1255 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 1256 |
+
"tokens/trainable": 40649
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 0.3515625,
|
| 1260 |
+
"grad_norm": 22.75,
|
| 1261 |
+
"learning_rate": 5e-06,
|
| 1262 |
+
"loss": 0.003507021814584732,
|
| 1263 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1264 |
+
"memory/max_active (GiB)": 46.9,
|
| 1265 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1266 |
+
"ppl": 1.00351,
|
| 1267 |
+
"step": 90,
|
| 1268 |
+
"tokens/total": 1961837,
|
| 1269 |
+
"tokens/train_per_sec_per_gpu": 3.85,
|
| 1270 |
+
"tokens/trainable": 41080
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 0.35546875,
|
| 1274 |
+
"grad_norm": 0.443359375,
|
| 1275 |
+
"learning_rate": 5e-06,
|
| 1276 |
+
"loss": 0.0008387751877307892,
|
| 1277 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1278 |
+
"memory/max_active (GiB)": 46.9,
|
| 1279 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1280 |
+
"ppl": 1.00084,
|
| 1281 |
+
"step": 91,
|
| 1282 |
+
"tokens/total": 1986102,
|
| 1283 |
+
"tokens/train_per_sec_per_gpu": 4.95,
|
| 1284 |
+
"tokens/trainable": 41599
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"epoch": 0.359375,
|
| 1288 |
+
"grad_norm": 2.75,
|
| 1289 |
+
"learning_rate": 5e-06,
|
| 1290 |
+
"loss": 0.0063427165150642395,
|
| 1291 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1292 |
+
"memory/max_active (GiB)": 46.9,
|
| 1293 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1294 |
+
"ppl": 1.00636,
|
| 1295 |
+
"step": 92,
|
| 1296 |
+
"tokens/total": 2008498,
|
| 1297 |
+
"tokens/train_per_sec_per_gpu": 3.79,
|
| 1298 |
+
"tokens/trainable": 42076
|
| 1299 |
+
},
|
| 1300 |
+
{
|
| 1301 |
+
"epoch": 0.36328125,
|
| 1302 |
+
"grad_norm": 8.375,
|
| 1303 |
+
"learning_rate": 5e-06,
|
| 1304 |
+
"loss": 0.0007749572396278381,
|
| 1305 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1306 |
+
"memory/max_active (GiB)": 46.9,
|
| 1307 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1308 |
+
"ppl": 1.00078,
|
| 1309 |
+
"step": 93,
|
| 1310 |
+
"tokens/total": 2030279,
|
| 1311 |
+
"tokens/train_per_sec_per_gpu": 5.07,
|
| 1312 |
+
"tokens/trainable": 42522
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"epoch": 0.3671875,
|
| 1316 |
+
"grad_norm": 4.90625,
|
| 1317 |
+
"learning_rate": 5e-06,
|
| 1318 |
+
"loss": 0.02601948380470276,
|
| 1319 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1320 |
+
"memory/max_active (GiB)": 46.9,
|
| 1321 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1322 |
+
"ppl": 1.02636,
|
| 1323 |
+
"step": 94,
|
| 1324 |
+
"tokens/total": 2051914,
|
| 1325 |
+
"tokens/train_per_sec_per_gpu": 5.33,
|
| 1326 |
+
"tokens/trainable": 42974
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"epoch": 0.37109375,
|
| 1330 |
+
"grad_norm": 4.71875,
|
| 1331 |
+
"learning_rate": 5e-06,
|
| 1332 |
+
"loss": 0.015122350305318832,
|
| 1333 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1334 |
+
"memory/max_active (GiB)": 46.9,
|
| 1335 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1336 |
+
"ppl": 1.01524,
|
| 1337 |
+
"step": 95,
|
| 1338 |
+
"tokens/total": 2071768,
|
| 1339 |
+
"tokens/train_per_sec_per_gpu": 3.93,
|
| 1340 |
+
"tokens/trainable": 43380
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"epoch": 0.375,
|
| 1344 |
+
"grad_norm": 11.0625,
|
| 1345 |
+
"learning_rate": 5e-06,
|
| 1346 |
+
"loss": 0.01082584261894226,
|
| 1347 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1348 |
+
"memory/max_active (GiB)": 46.9,
|
| 1349 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1350 |
+
"ppl": 1.01088,
|
| 1351 |
+
"step": 96,
|
| 1352 |
+
"tokens/total": 2094957,
|
| 1353 |
+
"tokens/train_per_sec_per_gpu": 3.8,
|
| 1354 |
+
"tokens/trainable": 43865
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 0.37890625,
|
| 1358 |
+
"grad_norm": 3.953125,
|
| 1359 |
+
"learning_rate": 5e-06,
|
| 1360 |
+
"loss": 0.008022107183933258,
|
| 1361 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1362 |
+
"memory/max_active (GiB)": 46.9,
|
| 1363 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1364 |
+
"ppl": 1.00805,
|
| 1365 |
+
"step": 97,
|
| 1366 |
+
"tokens/total": 2117603,
|
| 1367 |
+
"tokens/train_per_sec_per_gpu": 2.97,
|
| 1368 |
+
"tokens/trainable": 44328
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"epoch": 0.3828125,
|
| 1372 |
+
"grad_norm": 8.875,
|
| 1373 |
+
"learning_rate": 5e-06,
|
| 1374 |
+
"loss": 0.010948814451694489,
|
| 1375 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1376 |
+
"memory/max_active (GiB)": 46.9,
|
| 1377 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1378 |
+
"ppl": 1.01101,
|
| 1379 |
+
"step": 98,
|
| 1380 |
+
"tokens/total": 2138416,
|
| 1381 |
+
"tokens/train_per_sec_per_gpu": 4.0,
|
| 1382 |
+
"tokens/trainable": 44778
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 0.38671875,
|
| 1386 |
+
"grad_norm": 7.5,
|
| 1387 |
+
"learning_rate": 5e-06,
|
| 1388 |
+
"loss": 0.01736413687467575,
|
| 1389 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1390 |
+
"memory/max_active (GiB)": 46.9,
|
| 1391 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1392 |
+
"ppl": 1.01752,
|
| 1393 |
+
"step": 99,
|
| 1394 |
+
"tokens/total": 2160046,
|
| 1395 |
+
"tokens/train_per_sec_per_gpu": 5.63,
|
| 1396 |
+
"tokens/trainable": 45225
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 0.390625,
|
| 1400 |
+
"grad_norm": 1.828125,
|
| 1401 |
+
"learning_rate": 5e-06,
|
| 1402 |
+
"loss": 0.002016715705394745,
|
| 1403 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1404 |
+
"memory/max_active (GiB)": 46.9,
|
| 1405 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1406 |
+
"ppl": 1.00202,
|
| 1407 |
+
"step": 100,
|
| 1408 |
+
"tokens/total": 2181923,
|
| 1409 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 1410 |
+
"tokens/trainable": 45678
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"epoch": 0.39453125,
|
| 1414 |
+
"grad_norm": 1.25,
|
| 1415 |
+
"learning_rate": 5e-06,
|
| 1416 |
+
"loss": 0.0019299909472465515,
|
| 1417 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1418 |
+
"memory/max_active (GiB)": 46.9,
|
| 1419 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1420 |
+
"ppl": 1.00193,
|
| 1421 |
+
"step": 101,
|
| 1422 |
+
"tokens/total": 2203558,
|
| 1423 |
+
"tokens/train_per_sec_per_gpu": 4.39,
|
| 1424 |
+
"tokens/trainable": 46125
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 0.3984375,
|
| 1428 |
+
"grad_norm": 3.859375,
|
| 1429 |
+
"learning_rate": 5e-06,
|
| 1430 |
+
"loss": 0.004272259771823883,
|
| 1431 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1432 |
+
"memory/max_active (GiB)": 46.9,
|
| 1433 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1434 |
+
"ppl": 1.00428,
|
| 1435 |
+
"step": 102,
|
| 1436 |
+
"tokens/total": 2226903,
|
| 1437 |
+
"tokens/train_per_sec_per_gpu": 4.99,
|
| 1438 |
+
"tokens/trainable": 46612
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"epoch": 0.40234375,
|
| 1442 |
+
"grad_norm": 8.5625,
|
| 1443 |
+
"learning_rate": 5e-06,
|
| 1444 |
+
"loss": 0.005463987588882446,
|
| 1445 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1446 |
+
"memory/max_active (GiB)": 46.9,
|
| 1447 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1448 |
+
"ppl": 1.00548,
|
| 1449 |
+
"step": 103,
|
| 1450 |
+
"tokens/total": 2248454,
|
| 1451 |
+
"tokens/train_per_sec_per_gpu": 4.56,
|
| 1452 |
+
"tokens/trainable": 47059
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"epoch": 0.40625,
|
| 1456 |
+
"grad_norm": 28.25,
|
| 1457 |
+
"learning_rate": 5e-06,
|
| 1458 |
+
"loss": 0.029229726642370224,
|
| 1459 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1460 |
+
"memory/max_active (GiB)": 46.9,
|
| 1461 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1462 |
+
"ppl": 1.02966,
|
| 1463 |
+
"step": 104,
|
| 1464 |
+
"tokens/total": 2271600,
|
| 1465 |
+
"tokens/train_per_sec_per_gpu": 4.31,
|
| 1466 |
+
"tokens/trainable": 47536
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"epoch": 0.41015625,
|
| 1470 |
+
"grad_norm": 3.8125,
|
| 1471 |
+
"learning_rate": 5e-06,
|
| 1472 |
+
"loss": 0.002427205443382263,
|
| 1473 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1474 |
+
"memory/max_active (GiB)": 46.9,
|
| 1475 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1476 |
+
"ppl": 1.00243,
|
| 1477 |
+
"step": 105,
|
| 1478 |
+
"tokens/total": 2295278,
|
| 1479 |
+
"tokens/train_per_sec_per_gpu": 3.72,
|
| 1480 |
+
"tokens/trainable": 48037
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"epoch": 0.4140625,
|
| 1484 |
+
"grad_norm": 3.734375,
|
| 1485 |
+
"learning_rate": 5e-06,
|
| 1486 |
+
"loss": 0.0015581585466861725,
|
| 1487 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1488 |
+
"memory/max_active (GiB)": 46.9,
|
| 1489 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1490 |
+
"ppl": 1.00156,
|
| 1491 |
+
"step": 106,
|
| 1492 |
+
"tokens/total": 2317645,
|
| 1493 |
+
"tokens/train_per_sec_per_gpu": 4.93,
|
| 1494 |
+
"tokens/trainable": 48491
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"epoch": 0.41796875,
|
| 1498 |
+
"grad_norm": 17.125,
|
| 1499 |
+
"learning_rate": 5e-06,
|
| 1500 |
+
"loss": 0.023405246436595917,
|
| 1501 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1502 |
+
"memory/max_active (GiB)": 46.9,
|
| 1503 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1504 |
+
"ppl": 1.02368,
|
| 1505 |
+
"step": 107,
|
| 1506 |
+
"tokens/total": 2339279,
|
| 1507 |
+
"tokens/train_per_sec_per_gpu": 4.48,
|
| 1508 |
+
"tokens/trainable": 48940
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 0.421875,
|
| 1512 |
+
"grad_norm": 2.53125,
|
| 1513 |
+
"learning_rate": 5e-06,
|
| 1514 |
+
"loss": 0.0015513133257627487,
|
| 1515 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1516 |
+
"memory/max_active (GiB)": 46.9,
|
| 1517 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1518 |
+
"ppl": 1.00155,
|
| 1519 |
+
"step": 108,
|
| 1520 |
+
"tokens/total": 2360540,
|
| 1521 |
+
"tokens/train_per_sec_per_gpu": 3.87,
|
| 1522 |
+
"tokens/trainable": 49389
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"epoch": 0.42578125,
|
| 1526 |
+
"grad_norm": 12.875,
|
| 1527 |
+
"learning_rate": 5e-06,
|
| 1528 |
+
"loss": 0.031832728534936905,
|
| 1529 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1530 |
+
"memory/max_active (GiB)": 46.9,
|
| 1531 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1532 |
+
"ppl": 1.03234,
|
| 1533 |
+
"step": 109,
|
| 1534 |
+
"tokens/total": 2381374,
|
| 1535 |
+
"tokens/train_per_sec_per_gpu": 4.73,
|
| 1536 |
+
"tokens/trainable": 49838
|
| 1537 |
+
},
|
| 1538 |
+
{
|
| 1539 |
+
"epoch": 0.4296875,
|
| 1540 |
+
"grad_norm": 6.71875,
|
| 1541 |
+
"learning_rate": 5e-06,
|
| 1542 |
+
"loss": 0.002377200871706009,
|
| 1543 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1544 |
+
"memory/max_active (GiB)": 46.9,
|
| 1545 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1546 |
+
"ppl": 1.00238,
|
| 1547 |
+
"step": 110,
|
| 1548 |
+
"tokens/total": 2403924,
|
| 1549 |
+
"tokens/train_per_sec_per_gpu": 4.38,
|
| 1550 |
+
"tokens/trainable": 50310
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"epoch": 0.43359375,
|
| 1554 |
+
"grad_norm": 1.1875,
|
| 1555 |
+
"learning_rate": 5e-06,
|
| 1556 |
+
"loss": 0.0006611049175262451,
|
| 1557 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1558 |
+
"memory/max_active (GiB)": 46.9,
|
| 1559 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1560 |
+
"ppl": 1.00066,
|
| 1561 |
+
"step": 111,
|
| 1562 |
+
"tokens/total": 2424591,
|
| 1563 |
+
"tokens/train_per_sec_per_gpu": 3.89,
|
| 1564 |
+
"tokens/trainable": 50755
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"epoch": 0.4375,
|
| 1568 |
+
"grad_norm": 1.359375,
|
| 1569 |
+
"learning_rate": 5e-06,
|
| 1570 |
+
"loss": 0.0006331540644168854,
|
| 1571 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1572 |
+
"memory/max_active (GiB)": 46.9,
|
| 1573 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1574 |
+
"ppl": 1.00063,
|
| 1575 |
+
"step": 112,
|
| 1576 |
+
"tokens/total": 2446829,
|
| 1577 |
+
"tokens/train_per_sec_per_gpu": 4.36,
|
| 1578 |
+
"tokens/trainable": 51213
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"epoch": 0.44140625,
|
| 1582 |
+
"grad_norm": 0.1259765625,
|
| 1583 |
+
"learning_rate": 5e-06,
|
| 1584 |
+
"loss": 0.0002308972179889679,
|
| 1585 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1586 |
+
"memory/max_active (GiB)": 46.9,
|
| 1587 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1588 |
+
"ppl": 1.00023,
|
| 1589 |
+
"step": 113,
|
| 1590 |
+
"tokens/total": 2467554,
|
| 1591 |
+
"tokens/train_per_sec_per_gpu": 5.3,
|
| 1592 |
+
"tokens/trainable": 51660
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 0.4453125,
|
| 1596 |
+
"grad_norm": 0.53515625,
|
| 1597 |
+
"learning_rate": 5e-06,
|
| 1598 |
+
"loss": 0.00044569000601768494,
|
| 1599 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1600 |
+
"memory/max_active (GiB)": 46.9,
|
| 1601 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1602 |
+
"ppl": 1.00045,
|
| 1603 |
+
"step": 114,
|
| 1604 |
+
"tokens/total": 2490093,
|
| 1605 |
+
"tokens/train_per_sec_per_gpu": 3.86,
|
| 1606 |
+
"tokens/trainable": 52133
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"epoch": 0.44921875,
|
| 1610 |
+
"grad_norm": 0.640625,
|
| 1611 |
+
"learning_rate": 5e-06,
|
| 1612 |
+
"loss": 0.000523759052157402,
|
| 1613 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1614 |
+
"memory/max_active (GiB)": 46.9,
|
| 1615 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1616 |
+
"ppl": 1.00052,
|
| 1617 |
+
"step": 115,
|
| 1618 |
+
"tokens/total": 2511764,
|
| 1619 |
+
"tokens/train_per_sec_per_gpu": 5.14,
|
| 1620 |
+
"tokens/trainable": 52580
|
| 1621 |
+
},
|
| 1622 |
+
{
|
| 1623 |
+
"epoch": 0.453125,
|
| 1624 |
+
"grad_norm": 10.3125,
|
| 1625 |
+
"learning_rate": 5e-06,
|
| 1626 |
+
"loss": 0.010446945205330849,
|
| 1627 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1628 |
+
"memory/max_active (GiB)": 46.9,
|
| 1629 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1630 |
+
"ppl": 1.0105,
|
| 1631 |
+
"step": 116,
|
| 1632 |
+
"tokens/total": 2533541,
|
| 1633 |
+
"tokens/train_per_sec_per_gpu": 3.82,
|
| 1634 |
+
"tokens/trainable": 53033
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"epoch": 0.45703125,
|
| 1638 |
+
"grad_norm": 6.625,
|
| 1639 |
+
"learning_rate": 5e-06,
|
| 1640 |
+
"loss": 0.0046858396381139755,
|
| 1641 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1642 |
+
"memory/max_active (GiB)": 46.9,
|
| 1643 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1644 |
+
"ppl": 1.0047,
|
| 1645 |
+
"step": 117,
|
| 1646 |
+
"tokens/total": 2556046,
|
| 1647 |
+
"tokens/train_per_sec_per_gpu": 5.67,
|
| 1648 |
+
"tokens/trainable": 53517
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"epoch": 0.4609375,
|
| 1652 |
+
"grad_norm": 4.40625,
|
| 1653 |
+
"learning_rate": 5e-06,
|
| 1654 |
+
"loss": 0.006048301234841347,
|
| 1655 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1656 |
+
"memory/max_active (GiB)": 46.9,
|
| 1657 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1658 |
+
"ppl": 1.00607,
|
| 1659 |
+
"step": 118,
|
| 1660 |
+
"tokens/total": 2577387,
|
| 1661 |
+
"tokens/train_per_sec_per_gpu": 5.21,
|
| 1662 |
+
"tokens/trainable": 53967
|
| 1663 |
+
},
|
| 1664 |
+
{
|
| 1665 |
+
"epoch": 0.46484375,
|
| 1666 |
+
"grad_norm": 3.328125,
|
| 1667 |
+
"learning_rate": 5e-06,
|
| 1668 |
+
"loss": 0.005638599395751953,
|
| 1669 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 1670 |
+
"memory/max_active (GiB)": 46.9,
|
| 1671 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1672 |
+
"ppl": 1.00565,
|
| 1673 |
+
"step": 119,
|
| 1674 |
+
"tokens/total": 2599912,
|
| 1675 |
+
"tokens/train_per_sec_per_gpu": 5.04,
|
| 1676 |
+
"tokens/trainable": 54430
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 0.46875,
|
| 1680 |
+
"grad_norm": 12.5625,
|
| 1681 |
+
"learning_rate": 5e-06,
|
| 1682 |
+
"loss": 0.004470301792025566,
|
| 1683 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1684 |
+
"memory/max_active (GiB)": 46.9,
|
| 1685 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1686 |
+
"ppl": 1.00448,
|
| 1687 |
+
"step": 120,
|
| 1688 |
+
"tokens/total": 2622592,
|
| 1689 |
+
"tokens/train_per_sec_per_gpu": 5.57,
|
| 1690 |
+
"tokens/trainable": 54896
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"epoch": 0.47265625,
|
| 1694 |
+
"grad_norm": 10.25,
|
| 1695 |
+
"learning_rate": 5e-06,
|
| 1696 |
+
"loss": 0.007355847395956516,
|
| 1697 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1698 |
+
"memory/max_active (GiB)": 46.9,
|
| 1699 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1700 |
+
"ppl": 1.00738,
|
| 1701 |
+
"step": 121,
|
| 1702 |
+
"tokens/total": 2646024,
|
| 1703 |
+
"tokens/train_per_sec_per_gpu": 5.73,
|
| 1704 |
+
"tokens/trainable": 55379
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"epoch": 0.4765625,
|
| 1708 |
+
"grad_norm": 2.125,
|
| 1709 |
+
"learning_rate": 5e-06,
|
| 1710 |
+
"loss": 0.0003274921327829361,
|
| 1711 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1712 |
+
"memory/max_active (GiB)": 46.9,
|
| 1713 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1714 |
+
"ppl": 1.00033,
|
| 1715 |
+
"step": 122,
|
| 1716 |
+
"tokens/total": 2666034,
|
| 1717 |
+
"tokens/train_per_sec_per_gpu": 3.8,
|
| 1718 |
+
"tokens/trainable": 55803
|
| 1719 |
+
},
|
| 1720 |
+
{
|
| 1721 |
+
"epoch": 0.48046875,
|
| 1722 |
+
"grad_norm": 3.421875,
|
| 1723 |
+
"learning_rate": 5e-06,
|
| 1724 |
+
"loss": 0.0009368900209665298,
|
| 1725 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1726 |
+
"memory/max_active (GiB)": 46.9,
|
| 1727 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1728 |
+
"ppl": 1.00094,
|
| 1729 |
+
"step": 123,
|
| 1730 |
+
"tokens/total": 2687583,
|
| 1731 |
+
"tokens/train_per_sec_per_gpu": 3.76,
|
| 1732 |
+
"tokens/trainable": 56249
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"epoch": 0.484375,
|
| 1736 |
+
"grad_norm": 0.52734375,
|
| 1737 |
+
"learning_rate": 5e-06,
|
| 1738 |
+
"loss": 0.0003253612667322159,
|
| 1739 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1740 |
+
"memory/max_active (GiB)": 46.9,
|
| 1741 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1742 |
+
"ppl": 1.00033,
|
| 1743 |
+
"step": 124,
|
| 1744 |
+
"tokens/total": 2709310,
|
| 1745 |
+
"tokens/train_per_sec_per_gpu": 4.41,
|
| 1746 |
+
"tokens/trainable": 56713
|
| 1747 |
+
},
|
| 1748 |
+
{
|
| 1749 |
+
"epoch": 0.48828125,
|
| 1750 |
+
"grad_norm": 22.625,
|
| 1751 |
+
"learning_rate": 5e-06,
|
| 1752 |
+
"loss": 0.0033743418753147125,
|
| 1753 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1754 |
+
"memory/max_active (GiB)": 46.9,
|
| 1755 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1756 |
+
"ppl": 1.00338,
|
| 1757 |
+
"step": 125,
|
| 1758 |
+
"tokens/total": 2731725,
|
| 1759 |
+
"tokens/train_per_sec_per_gpu": 5.25,
|
| 1760 |
+
"tokens/trainable": 57196
|
| 1761 |
+
},
|
| 1762 |
+
{
|
| 1763 |
+
"epoch": 0.4921875,
|
| 1764 |
+
"grad_norm": 23.625,
|
| 1765 |
+
"learning_rate": 5e-06,
|
| 1766 |
+
"loss": 0.005839193239808083,
|
| 1767 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1768 |
+
"memory/max_active (GiB)": 46.9,
|
| 1769 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1770 |
+
"ppl": 1.00586,
|
| 1771 |
+
"step": 126,
|
| 1772 |
+
"tokens/total": 2754818,
|
| 1773 |
+
"tokens/train_per_sec_per_gpu": 5.01,
|
| 1774 |
+
"tokens/trainable": 57680
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"epoch": 0.49609375,
|
| 1778 |
+
"grad_norm": 1.953125,
|
| 1779 |
+
"learning_rate": 5e-06,
|
| 1780 |
+
"loss": 0.0013373112305998802,
|
| 1781 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1782 |
+
"memory/max_active (GiB)": 46.9,
|
| 1783 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1784 |
+
"ppl": 1.00134,
|
| 1785 |
+
"step": 127,
|
| 1786 |
+
"tokens/total": 2775918,
|
| 1787 |
+
"tokens/train_per_sec_per_gpu": 4.39,
|
| 1788 |
+
"tokens/trainable": 58110
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"epoch": 0.5,
|
| 1792 |
+
"grad_norm": 5.90625,
|
| 1793 |
+
"learning_rate": 5e-06,
|
| 1794 |
+
"loss": 0.002514144405722618,
|
| 1795 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 1796 |
+
"memory/max_active (GiB)": 46.9,
|
| 1797 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 1798 |
+
"ppl": 1.00252,
|
| 1799 |
+
"step": 128,
|
| 1800 |
+
"tokens/total": 2796262,
|
| 1801 |
+
"tokens/train_per_sec_per_gpu": 4.5,
|
| 1802 |
+
"tokens/trainable": 58534
|
| 1803 |
+
}
|
| 1804 |
+
],
|
| 1805 |
+
"logging_steps": 1,
|
| 1806 |
+
"max_steps": 512,
|
| 1807 |
+
"num_input_tokens_seen": 0,
|
| 1808 |
+
"num_train_epochs": 2,
|
| 1809 |
+
"save_steps": 500,
|
| 1810 |
+
"stateful_callbacks": {
|
| 1811 |
+
"TrainerControl": {
|
| 1812 |
+
"args": {
|
| 1813 |
+
"should_epoch_stop": false,
|
| 1814 |
+
"should_evaluate": false,
|
| 1815 |
+
"should_log": false,
|
| 1816 |
+
"should_save": true,
|
| 1817 |
+
"should_training_stop": false
|
| 1818 |
+
},
|
| 1819 |
+
"attributes": {}
|
| 1820 |
+
}
|
| 1821 |
+
},
|
| 1822 |
+
"total_flos": 1.8750157626330317e+17,
|
| 1823 |
+
"train_batch_size": 1,
|
| 1824 |
+
"trial_name": null,
|
| 1825 |
+
"trial_params": null
|
| 1826 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-128/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0107cd88280e968b63f91f7125176eba0c9002c1f534529b1c400b807d0860f
|
| 3 |
+
size 8913
|
full_aft_midtrain4/balanced/checkpoint-16/chat_template.jinja
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{ bos_token }}
|
| 2 |
+
{%- if messages[0]['role'] == 'system' -%}
|
| 3 |
+
{%- if messages[0]['content'] is string -%}
|
| 4 |
+
{%- set first_user_prefix = messages[0]['content'] + '
|
| 5 |
+
|
| 6 |
+
' -%}
|
| 7 |
+
{%- else -%}
|
| 8 |
+
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
|
| 9 |
+
|
| 10 |
+
' -%}
|
| 11 |
+
{%- endif -%}
|
| 12 |
+
{%- set loop_messages = messages[1:] -%}
|
| 13 |
+
{%- else -%}
|
| 14 |
+
{%- set first_user_prefix = "" -%}
|
| 15 |
+
{%- set loop_messages = messages -%}
|
| 16 |
+
{%- endif -%}
|
| 17 |
+
{%- for message in loop_messages -%}
|
| 18 |
+
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
|
| 19 |
+
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if (message['role'] == 'assistant') -%}
|
| 22 |
+
{%- set role = "model" -%}
|
| 23 |
+
{%- else -%}
|
| 24 |
+
{%- set role = message['role'] -%}
|
| 25 |
+
{%- endif -%}
|
| 26 |
+
{{ '<start_of_turn>' + role + '
|
| 27 |
+
' + (first_user_prefix if loop.first else "") }}
|
| 28 |
+
{%- if message['content'] is string -%}
|
| 29 |
+
{{ message['content'] | trim }}
|
| 30 |
+
{%- elif message['content'] is iterable -%}
|
| 31 |
+
{%- for item in message['content'] -%}
|
| 32 |
+
{%- if item['type'] == 'image' -%}
|
| 33 |
+
{{ '<start_of_image>' }}
|
| 34 |
+
{%- elif item['type'] == 'text' -%}
|
| 35 |
+
{{ item['text'] | trim }}
|
| 36 |
+
{%- endif -%}
|
| 37 |
+
{%- endfor -%}
|
| 38 |
+
{%- else -%}
|
| 39 |
+
{{ raise_exception("Invalid content type") }}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{ '<end_of_turn>
|
| 42 |
+
' }}
|
| 43 |
+
{%- endfor -%}
|
| 44 |
+
{%- if add_generation_prompt -%}
|
| 45 |
+
{{'<start_of_turn>model
|
| 46 |
+
'}}
|
| 47 |
+
{%- endif -%}
|
full_aft_midtrain4/balanced/checkpoint-16/config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Gemma3ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 255999,
|
| 6 |
+
"bos_token_id": 2,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eoi_token_index": 256000,
|
| 9 |
+
"eos_token_id": 1,
|
| 10 |
+
"image_token_index": 262144,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"mm_tokens_per_image": 256,
|
| 13 |
+
"model_type": "gemma3",
|
| 14 |
+
"pad_token_id": 0,
|
| 15 |
+
"text_config": {
|
| 16 |
+
"_sliding_window_pattern": 6,
|
| 17 |
+
"attention_bias": false,
|
| 18 |
+
"attention_dropout": 0.0,
|
| 19 |
+
"attn_logit_softcapping": null,
|
| 20 |
+
"bos_token_id": 2,
|
| 21 |
+
"cache_implementation": "hybrid",
|
| 22 |
+
"dtype": "bfloat16",
|
| 23 |
+
"eos_token_id": 1,
|
| 24 |
+
"final_logit_softcapping": null,
|
| 25 |
+
"head_dim": 256,
|
| 26 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 27 |
+
"hidden_size": 3840,
|
| 28 |
+
"initializer_range": 0.02,
|
| 29 |
+
"intermediate_size": 15360,
|
| 30 |
+
"layer_types": [
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"sliding_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"sliding_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"sliding_attention",
|
| 59 |
+
"sliding_attention",
|
| 60 |
+
"full_attention",
|
| 61 |
+
"sliding_attention",
|
| 62 |
+
"sliding_attention",
|
| 63 |
+
"sliding_attention",
|
| 64 |
+
"sliding_attention",
|
| 65 |
+
"sliding_attention",
|
| 66 |
+
"full_attention",
|
| 67 |
+
"sliding_attention",
|
| 68 |
+
"sliding_attention",
|
| 69 |
+
"sliding_attention",
|
| 70 |
+
"sliding_attention",
|
| 71 |
+
"sliding_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"sliding_attention",
|
| 74 |
+
"sliding_attention",
|
| 75 |
+
"sliding_attention",
|
| 76 |
+
"sliding_attention",
|
| 77 |
+
"sliding_attention",
|
| 78 |
+
"full_attention"
|
| 79 |
+
],
|
| 80 |
+
"max_position_embeddings": 131072,
|
| 81 |
+
"model_type": "gemma3_text",
|
| 82 |
+
"num_attention_heads": 16,
|
| 83 |
+
"num_hidden_layers": 48,
|
| 84 |
+
"num_key_value_heads": 8,
|
| 85 |
+
"pad_token_id": 0,
|
| 86 |
+
"query_pre_attn_scalar": 256,
|
| 87 |
+
"rms_norm_eps": 1e-06,
|
| 88 |
+
"rope_parameters": {
|
| 89 |
+
"full_attention": {
|
| 90 |
+
"factor": 8.0,
|
| 91 |
+
"rope_theta": 1000000.0,
|
| 92 |
+
"rope_type": "linear"
|
| 93 |
+
},
|
| 94 |
+
"sliding_attention": {
|
| 95 |
+
"rope_theta": 10000.0,
|
| 96 |
+
"rope_type": "default"
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"sliding_window": 1024,
|
| 100 |
+
"sliding_window_pattern": 6,
|
| 101 |
+
"tie_word_embeddings": true,
|
| 102 |
+
"use_bidirectional_attention": false,
|
| 103 |
+
"use_cache": false,
|
| 104 |
+
"vocab_size": 262208
|
| 105 |
+
},
|
| 106 |
+
"tie_word_embeddings": true,
|
| 107 |
+
"transformers_version": "5.9.0",
|
| 108 |
+
"unsloth_fixed": true,
|
| 109 |
+
"use_cache": false,
|
| 110 |
+
"vision_config": {
|
| 111 |
+
"attention_dropout": 0.0,
|
| 112 |
+
"dtype": "bfloat16",
|
| 113 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 114 |
+
"hidden_size": 1152,
|
| 115 |
+
"image_size": 896,
|
| 116 |
+
"intermediate_size": 4304,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"model_type": "siglip_vision_model",
|
| 119 |
+
"num_attention_heads": 16,
|
| 120 |
+
"num_channels": 3,
|
| 121 |
+
"num_hidden_layers": 27,
|
| 122 |
+
"patch_size": 14,
|
| 123 |
+
"vision_use_head": false
|
| 124 |
+
}
|
| 125 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 2,
|
| 3 |
+
"cache_implementation": "hybrid",
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
1,
|
| 7 |
+
106
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 0,
|
| 10 |
+
"top_k": 64,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.9.0"
|
| 13 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:002935554ab69eb9b7e6de5d0d95d0a99a92cd58095478eb7c0861022d0dc671
|
| 3 |
+
size 26388552360
|
full_aft_midtrain4/balanced/checkpoint-16/preprocessor_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pan_and_scan": null,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": true,
|
| 7 |
+
"image_mean": [
|
| 8 |
+
0.5,
|
| 9 |
+
0.5,
|
| 10 |
+
0.5
|
| 11 |
+
],
|
| 12 |
+
"image_processor_type": "Gemma3ImageProcessor",
|
| 13 |
+
"image_seq_length": 256,
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.5,
|
| 16 |
+
0.5,
|
| 17 |
+
0.5
|
| 18 |
+
],
|
| 19 |
+
"pan_and_scan_max_num_crops": null,
|
| 20 |
+
"pan_and_scan_min_crop_size": null,
|
| 21 |
+
"pan_and_scan_min_ratio_to_activate": null,
|
| 22 |
+
"processor_class": "Gemma3Processor",
|
| 23 |
+
"resample": 2,
|
| 24 |
+
"rescale_factor": 0.00392156862745098,
|
| 25 |
+
"size": {
|
| 26 |
+
"height": 896,
|
| 27 |
+
"width": 896
|
| 28 |
+
}
|
| 29 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/processor_config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_seq_length": 256,
|
| 3 |
+
"processor_class": "Gemma3Processor"
|
| 4 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
|
| 3 |
+
size 33384567
|
full_aft_midtrain4/balanced/checkpoint-16/tokenizer_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"boi_token": "<start_of_image>",
|
| 4 |
+
"bos_token": "<bos>",
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eoi_token": "<end_of_image>",
|
| 7 |
+
"eos_token": "<eos>",
|
| 8 |
+
"image_token": "<image_soft_token>",
|
| 9 |
+
"is_local": false,
|
| 10 |
+
"local_files_only": false,
|
| 11 |
+
"mask_token": "<mask>",
|
| 12 |
+
"model_max_length": 131072,
|
| 13 |
+
"model_specific_special_tokens": {
|
| 14 |
+
"boi_token": "<start_of_image>",
|
| 15 |
+
"eoi_token": "<end_of_image>",
|
| 16 |
+
"image_token": "<image_soft_token>"
|
| 17 |
+
},
|
| 18 |
+
"pad_token": "<pad>",
|
| 19 |
+
"padding_side": "left",
|
| 20 |
+
"processor_class": "Gemma3Processor",
|
| 21 |
+
"sp_model_kwargs": null,
|
| 22 |
+
"spaces_between_special_tokens": false,
|
| 23 |
+
"tokenizer_class": "GemmaTokenizer",
|
| 24 |
+
"unk_token": "<unk>",
|
| 25 |
+
"use_default_system_prompt": false
|
| 26 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/tokens_state.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"total": 347838, "trainable": 7258}
|
full_aft_midtrain4/balanced/checkpoint-16/trainer_state.json
ADDED
|
@@ -0,0 +1,258 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.0625,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 16,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.00390625,
|
| 14 |
+
"grad_norm": 17.375,
|
| 15 |
+
"learning_rate": 5e-06,
|
| 16 |
+
"loss": 0.11627960205078125,
|
| 17 |
+
"memory/device_reserved (GiB)": 46.13,
|
| 18 |
+
"memory/max_active (GiB)": 35.94,
|
| 19 |
+
"memory/max_allocated (GiB)": 35.94,
|
| 20 |
+
"ppl": 1.12331,
|
| 21 |
+
"step": 1,
|
| 22 |
+
"tokens/total": 22614,
|
| 23 |
+
"tokens/train_per_sec_per_gpu": 1.99,
|
| 24 |
+
"tokens/trainable": 470
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.0078125,
|
| 28 |
+
"grad_norm": 1472.0,
|
| 29 |
+
"learning_rate": 5e-06,
|
| 30 |
+
"loss": 0.1474212110042572,
|
| 31 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 32 |
+
"memory/max_active (GiB)": 46.9,
|
| 33 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 34 |
+
"ppl": 1.15884,
|
| 35 |
+
"step": 2,
|
| 36 |
+
"tokens/total": 44116,
|
| 37 |
+
"tokens/train_per_sec_per_gpu": 4.14,
|
| 38 |
+
"tokens/trainable": 922
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.01171875,
|
| 42 |
+
"grad_norm": 57.75,
|
| 43 |
+
"learning_rate": 5e-06,
|
| 44 |
+
"loss": 0.2154746651649475,
|
| 45 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 46 |
+
"memory/max_active (GiB)": 46.9,
|
| 47 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 48 |
+
"ppl": 1.24045,
|
| 49 |
+
"step": 3,
|
| 50 |
+
"tokens/total": 67092,
|
| 51 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 52 |
+
"tokens/trainable": 1396
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"epoch": 0.015625,
|
| 56 |
+
"grad_norm": 37.75,
|
| 57 |
+
"learning_rate": 5e-06,
|
| 58 |
+
"loss": 0.10307750105857849,
|
| 59 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 60 |
+
"memory/max_active (GiB)": 46.9,
|
| 61 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 62 |
+
"ppl": 1.10858,
|
| 63 |
+
"step": 4,
|
| 64 |
+
"tokens/total": 88800,
|
| 65 |
+
"tokens/train_per_sec_per_gpu": 5.16,
|
| 66 |
+
"tokens/trainable": 1850
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"epoch": 0.01953125,
|
| 70 |
+
"grad_norm": 25.0,
|
| 71 |
+
"learning_rate": 5e-06,
|
| 72 |
+
"loss": 0.10710409283638,
|
| 73 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 74 |
+
"memory/max_active (GiB)": 46.9,
|
| 75 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 76 |
+
"ppl": 1.11305,
|
| 77 |
+
"step": 5,
|
| 78 |
+
"tokens/total": 110742,
|
| 79 |
+
"tokens/train_per_sec_per_gpu": 3.92,
|
| 80 |
+
"tokens/trainable": 2302
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.0234375,
|
| 84 |
+
"grad_norm": 16.25,
|
| 85 |
+
"learning_rate": 5e-06,
|
| 86 |
+
"loss": 0.05220422148704529,
|
| 87 |
+
"memory/device_reserved (GiB)": 57.19,
|
| 88 |
+
"memory/max_active (GiB)": 46.9,
|
| 89 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 90 |
+
"ppl": 1.05359,
|
| 91 |
+
"step": 6,
|
| 92 |
+
"tokens/total": 131497,
|
| 93 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 94 |
+
"tokens/trainable": 2742
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"epoch": 0.02734375,
|
| 98 |
+
"grad_norm": 18.0,
|
| 99 |
+
"learning_rate": 5e-06,
|
| 100 |
+
"loss": 0.06188321113586426,
|
| 101 |
+
"memory/device_reserved (GiB)": 59.04,
|
| 102 |
+
"memory/max_active (GiB)": 46.9,
|
| 103 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 104 |
+
"ppl": 1.06384,
|
| 105 |
+
"step": 7,
|
| 106 |
+
"tokens/total": 153846,
|
| 107 |
+
"tokens/train_per_sec_per_gpu": 4.99,
|
| 108 |
+
"tokens/trainable": 3208
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 0.03125,
|
| 112 |
+
"grad_norm": 16.75,
|
| 113 |
+
"learning_rate": 5e-06,
|
| 114 |
+
"loss": 0.036078453063964844,
|
| 115 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 116 |
+
"memory/max_active (GiB)": 46.9,
|
| 117 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 118 |
+
"ppl": 1.03674,
|
| 119 |
+
"step": 8,
|
| 120 |
+
"tokens/total": 175713,
|
| 121 |
+
"tokens/train_per_sec_per_gpu": 4.47,
|
| 122 |
+
"tokens/trainable": 3655
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.03515625,
|
| 126 |
+
"grad_norm": 13.5625,
|
| 127 |
+
"learning_rate": 5e-06,
|
| 128 |
+
"loss": 0.05783689022064209,
|
| 129 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 130 |
+
"memory/max_active (GiB)": 46.9,
|
| 131 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 132 |
+
"ppl": 1.05954,
|
| 133 |
+
"step": 9,
|
| 134 |
+
"tokens/total": 196358,
|
| 135 |
+
"tokens/train_per_sec_per_gpu": 3.39,
|
| 136 |
+
"tokens/trainable": 4091
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"epoch": 0.0390625,
|
| 140 |
+
"grad_norm": 16.375,
|
| 141 |
+
"learning_rate": 5e-06,
|
| 142 |
+
"loss": 0.05202341079711914,
|
| 143 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 144 |
+
"memory/max_active (GiB)": 46.9,
|
| 145 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 146 |
+
"ppl": 1.0534,
|
| 147 |
+
"step": 10,
|
| 148 |
+
"tokens/total": 218622,
|
| 149 |
+
"tokens/train_per_sec_per_gpu": 4.33,
|
| 150 |
+
"tokens/trainable": 4553
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 0.04296875,
|
| 154 |
+
"grad_norm": 22.5,
|
| 155 |
+
"learning_rate": 5e-06,
|
| 156 |
+
"loss": 0.03920435905456543,
|
| 157 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 158 |
+
"memory/max_active (GiB)": 46.9,
|
| 159 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 160 |
+
"ppl": 1.03998,
|
| 161 |
+
"step": 11,
|
| 162 |
+
"tokens/total": 239835,
|
| 163 |
+
"tokens/train_per_sec_per_gpu": 4.54,
|
| 164 |
+
"tokens/trainable": 4992
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.046875,
|
| 168 |
+
"grad_norm": 43.75,
|
| 169 |
+
"learning_rate": 5e-06,
|
| 170 |
+
"loss": 0.0486750602722168,
|
| 171 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 172 |
+
"memory/max_active (GiB)": 46.9,
|
| 173 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 174 |
+
"ppl": 1.04988,
|
| 175 |
+
"step": 12,
|
| 176 |
+
"tokens/total": 264428,
|
| 177 |
+
"tokens/train_per_sec_per_gpu": 4.93,
|
| 178 |
+
"tokens/trainable": 5494
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"epoch": 0.05078125,
|
| 182 |
+
"grad_norm": 7.59375,
|
| 183 |
+
"learning_rate": 5e-06,
|
| 184 |
+
"loss": 0.023757457733154297,
|
| 185 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 186 |
+
"memory/max_active (GiB)": 46.9,
|
| 187 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 188 |
+
"ppl": 1.02404,
|
| 189 |
+
"step": 13,
|
| 190 |
+
"tokens/total": 284859,
|
| 191 |
+
"tokens/train_per_sec_per_gpu": 3.95,
|
| 192 |
+
"tokens/trainable": 5933
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"epoch": 0.0546875,
|
| 196 |
+
"grad_norm": 19.0,
|
| 197 |
+
"learning_rate": 5e-06,
|
| 198 |
+
"loss": 0.04330146312713623,
|
| 199 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 200 |
+
"memory/max_active (GiB)": 46.9,
|
| 201 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 202 |
+
"ppl": 1.04425,
|
| 203 |
+
"step": 14,
|
| 204 |
+
"tokens/total": 305640,
|
| 205 |
+
"tokens/train_per_sec_per_gpu": 3.17,
|
| 206 |
+
"tokens/trainable": 6369
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.05859375,
|
| 210 |
+
"grad_norm": 9.6875,
|
| 211 |
+
"learning_rate": 5e-06,
|
| 212 |
+
"loss": 0.03816962242126465,
|
| 213 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 214 |
+
"memory/max_active (GiB)": 46.9,
|
| 215 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 216 |
+
"ppl": 1.03891,
|
| 217 |
+
"step": 15,
|
| 218 |
+
"tokens/total": 327063,
|
| 219 |
+
"tokens/train_per_sec_per_gpu": 4.35,
|
| 220 |
+
"tokens/trainable": 6820
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.0625,
|
| 224 |
+
"grad_norm": 11.8125,
|
| 225 |
+
"learning_rate": 5e-06,
|
| 226 |
+
"loss": 0.03375864028930664,
|
| 227 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 228 |
+
"memory/max_active (GiB)": 46.9,
|
| 229 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 230 |
+
"ppl": 1.03433,
|
| 231 |
+
"step": 16,
|
| 232 |
+
"tokens/total": 347838,
|
| 233 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 234 |
+
"tokens/trainable": 7258
|
| 235 |
+
}
|
| 236 |
+
],
|
| 237 |
+
"logging_steps": 1,
|
| 238 |
+
"max_steps": 512,
|
| 239 |
+
"num_input_tokens_seen": 0,
|
| 240 |
+
"num_train_epochs": 2,
|
| 241 |
+
"save_steps": 500,
|
| 242 |
+
"stateful_callbacks": {
|
| 243 |
+
"TrainerControl": {
|
| 244 |
+
"args": {
|
| 245 |
+
"should_epoch_stop": false,
|
| 246 |
+
"should_evaluate": false,
|
| 247 |
+
"should_log": false,
|
| 248 |
+
"should_save": true,
|
| 249 |
+
"should_training_stop": false
|
| 250 |
+
},
|
| 251 |
+
"attributes": {}
|
| 252 |
+
}
|
| 253 |
+
},
|
| 254 |
+
"total_flos": 2.332405666100019e+16,
|
| 255 |
+
"train_batch_size": 1,
|
| 256 |
+
"trial_name": null,
|
| 257 |
+
"trial_params": null
|
| 258 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-16/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0107cd88280e968b63f91f7125176eba0c9002c1f534529b1c400b807d0860f
|
| 3 |
+
size 8913
|
full_aft_midtrain4/balanced/checkpoint-256/chat_template.jinja
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{ bos_token }}
|
| 2 |
+
{%- if messages[0]['role'] == 'system' -%}
|
| 3 |
+
{%- if messages[0]['content'] is string -%}
|
| 4 |
+
{%- set first_user_prefix = messages[0]['content'] + '
|
| 5 |
+
|
| 6 |
+
' -%}
|
| 7 |
+
{%- else -%}
|
| 8 |
+
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
|
| 9 |
+
|
| 10 |
+
' -%}
|
| 11 |
+
{%- endif -%}
|
| 12 |
+
{%- set loop_messages = messages[1:] -%}
|
| 13 |
+
{%- else -%}
|
| 14 |
+
{%- set first_user_prefix = "" -%}
|
| 15 |
+
{%- set loop_messages = messages -%}
|
| 16 |
+
{%- endif -%}
|
| 17 |
+
{%- for message in loop_messages -%}
|
| 18 |
+
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
|
| 19 |
+
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if (message['role'] == 'assistant') -%}
|
| 22 |
+
{%- set role = "model" -%}
|
| 23 |
+
{%- else -%}
|
| 24 |
+
{%- set role = message['role'] -%}
|
| 25 |
+
{%- endif -%}
|
| 26 |
+
{{ '<start_of_turn>' + role + '
|
| 27 |
+
' + (first_user_prefix if loop.first else "") }}
|
| 28 |
+
{%- if message['content'] is string -%}
|
| 29 |
+
{{ message['content'] | trim }}
|
| 30 |
+
{%- elif message['content'] is iterable -%}
|
| 31 |
+
{%- for item in message['content'] -%}
|
| 32 |
+
{%- if item['type'] == 'image' -%}
|
| 33 |
+
{{ '<start_of_image>' }}
|
| 34 |
+
{%- elif item['type'] == 'text' -%}
|
| 35 |
+
{{ item['text'] | trim }}
|
| 36 |
+
{%- endif -%}
|
| 37 |
+
{%- endfor -%}
|
| 38 |
+
{%- else -%}
|
| 39 |
+
{{ raise_exception("Invalid content type") }}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{ '<end_of_turn>
|
| 42 |
+
' }}
|
| 43 |
+
{%- endfor -%}
|
| 44 |
+
{%- if add_generation_prompt -%}
|
| 45 |
+
{{'<start_of_turn>model
|
| 46 |
+
'}}
|
| 47 |
+
{%- endif -%}
|
full_aft_midtrain4/balanced/checkpoint-256/config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Gemma3ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 255999,
|
| 6 |
+
"bos_token_id": 2,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eoi_token_index": 256000,
|
| 9 |
+
"eos_token_id": 1,
|
| 10 |
+
"image_token_index": 262144,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"mm_tokens_per_image": 256,
|
| 13 |
+
"model_type": "gemma3",
|
| 14 |
+
"pad_token_id": 0,
|
| 15 |
+
"text_config": {
|
| 16 |
+
"_sliding_window_pattern": 6,
|
| 17 |
+
"attention_bias": false,
|
| 18 |
+
"attention_dropout": 0.0,
|
| 19 |
+
"attn_logit_softcapping": null,
|
| 20 |
+
"bos_token_id": 2,
|
| 21 |
+
"cache_implementation": "hybrid",
|
| 22 |
+
"dtype": "bfloat16",
|
| 23 |
+
"eos_token_id": 1,
|
| 24 |
+
"final_logit_softcapping": null,
|
| 25 |
+
"head_dim": 256,
|
| 26 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 27 |
+
"hidden_size": 3840,
|
| 28 |
+
"initializer_range": 0.02,
|
| 29 |
+
"intermediate_size": 15360,
|
| 30 |
+
"layer_types": [
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"sliding_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"sliding_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"sliding_attention",
|
| 59 |
+
"sliding_attention",
|
| 60 |
+
"full_attention",
|
| 61 |
+
"sliding_attention",
|
| 62 |
+
"sliding_attention",
|
| 63 |
+
"sliding_attention",
|
| 64 |
+
"sliding_attention",
|
| 65 |
+
"sliding_attention",
|
| 66 |
+
"full_attention",
|
| 67 |
+
"sliding_attention",
|
| 68 |
+
"sliding_attention",
|
| 69 |
+
"sliding_attention",
|
| 70 |
+
"sliding_attention",
|
| 71 |
+
"sliding_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"sliding_attention",
|
| 74 |
+
"sliding_attention",
|
| 75 |
+
"sliding_attention",
|
| 76 |
+
"sliding_attention",
|
| 77 |
+
"sliding_attention",
|
| 78 |
+
"full_attention"
|
| 79 |
+
],
|
| 80 |
+
"max_position_embeddings": 131072,
|
| 81 |
+
"model_type": "gemma3_text",
|
| 82 |
+
"num_attention_heads": 16,
|
| 83 |
+
"num_hidden_layers": 48,
|
| 84 |
+
"num_key_value_heads": 8,
|
| 85 |
+
"pad_token_id": 0,
|
| 86 |
+
"query_pre_attn_scalar": 256,
|
| 87 |
+
"rms_norm_eps": 1e-06,
|
| 88 |
+
"rope_parameters": {
|
| 89 |
+
"full_attention": {
|
| 90 |
+
"factor": 8.0,
|
| 91 |
+
"rope_theta": 1000000.0,
|
| 92 |
+
"rope_type": "linear"
|
| 93 |
+
},
|
| 94 |
+
"sliding_attention": {
|
| 95 |
+
"rope_theta": 10000.0,
|
| 96 |
+
"rope_type": "default"
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"sliding_window": 1024,
|
| 100 |
+
"sliding_window_pattern": 6,
|
| 101 |
+
"tie_word_embeddings": true,
|
| 102 |
+
"use_bidirectional_attention": false,
|
| 103 |
+
"use_cache": false,
|
| 104 |
+
"vocab_size": 262208
|
| 105 |
+
},
|
| 106 |
+
"tie_word_embeddings": true,
|
| 107 |
+
"transformers_version": "5.9.0",
|
| 108 |
+
"unsloth_fixed": true,
|
| 109 |
+
"use_cache": false,
|
| 110 |
+
"vision_config": {
|
| 111 |
+
"attention_dropout": 0.0,
|
| 112 |
+
"dtype": "bfloat16",
|
| 113 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 114 |
+
"hidden_size": 1152,
|
| 115 |
+
"image_size": 896,
|
| 116 |
+
"intermediate_size": 4304,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"model_type": "siglip_vision_model",
|
| 119 |
+
"num_attention_heads": 16,
|
| 120 |
+
"num_channels": 3,
|
| 121 |
+
"num_hidden_layers": 27,
|
| 122 |
+
"patch_size": 14,
|
| 123 |
+
"vision_use_head": false
|
| 124 |
+
}
|
| 125 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-256/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 2,
|
| 3 |
+
"cache_implementation": "hybrid",
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
1,
|
| 7 |
+
106
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 0,
|
| 10 |
+
"top_k": 64,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.9.0"
|
| 13 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-256/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0dda9059ae44de482f67e01b52924e31cb4c5bb5646c6a1152be142261507ece
|
| 3 |
+
size 26388552360
|
full_aft_midtrain4/balanced/checkpoint-256/preprocessor_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pan_and_scan": null,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": true,
|
| 7 |
+
"image_mean": [
|
| 8 |
+
0.5,
|
| 9 |
+
0.5,
|
| 10 |
+
0.5
|
| 11 |
+
],
|
| 12 |
+
"image_processor_type": "Gemma3ImageProcessor",
|
| 13 |
+
"image_seq_length": 256,
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.5,
|
| 16 |
+
0.5,
|
| 17 |
+
0.5
|
| 18 |
+
],
|
| 19 |
+
"pan_and_scan_max_num_crops": null,
|
| 20 |
+
"pan_and_scan_min_crop_size": null,
|
| 21 |
+
"pan_and_scan_min_ratio_to_activate": null,
|
| 22 |
+
"processor_class": "Gemma3Processor",
|
| 23 |
+
"resample": 2,
|
| 24 |
+
"rescale_factor": 0.00392156862745098,
|
| 25 |
+
"size": {
|
| 26 |
+
"height": 896,
|
| 27 |
+
"width": 896
|
| 28 |
+
}
|
| 29 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-256/processor_config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_seq_length": 256,
|
| 3 |
+
"processor_class": "Gemma3Processor"
|
| 4 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-256/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
|
| 3 |
+
size 33384567
|
full_aft_midtrain4/balanced/checkpoint-256/tokenizer_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"boi_token": "<start_of_image>",
|
| 4 |
+
"bos_token": "<bos>",
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eoi_token": "<end_of_image>",
|
| 7 |
+
"eos_token": "<eos>",
|
| 8 |
+
"image_token": "<image_soft_token>",
|
| 9 |
+
"is_local": false,
|
| 10 |
+
"local_files_only": false,
|
| 11 |
+
"mask_token": "<mask>",
|
| 12 |
+
"model_max_length": 131072,
|
| 13 |
+
"model_specific_special_tokens": {
|
| 14 |
+
"boi_token": "<start_of_image>",
|
| 15 |
+
"eoi_token": "<end_of_image>",
|
| 16 |
+
"image_token": "<image_soft_token>"
|
| 17 |
+
},
|
| 18 |
+
"pad_token": "<pad>",
|
| 19 |
+
"padding_side": "left",
|
| 20 |
+
"processor_class": "Gemma3Processor",
|
| 21 |
+
"sp_model_kwargs": null,
|
| 22 |
+
"spaces_between_special_tokens": false,
|
| 23 |
+
"tokenizer_class": "GemmaTokenizer",
|
| 24 |
+
"unk_token": "<unk>",
|
| 25 |
+
"use_default_system_prompt": false
|
| 26 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-256/tokens_state.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"total": 5602336, "trainable": 117030}
|
full_aft_midtrain4/balanced/checkpoint-256/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
full_aft_midtrain4/balanced/checkpoint-256/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0107cd88280e968b63f91f7125176eba0c9002c1f534529b1c400b807d0860f
|
| 3 |
+
size 8913
|
full_aft_midtrain4/balanced/checkpoint-32/chat_template.jinja
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{ bos_token }}
|
| 2 |
+
{%- if messages[0]['role'] == 'system' -%}
|
| 3 |
+
{%- if messages[0]['content'] is string -%}
|
| 4 |
+
{%- set first_user_prefix = messages[0]['content'] + '
|
| 5 |
+
|
| 6 |
+
' -%}
|
| 7 |
+
{%- else -%}
|
| 8 |
+
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
|
| 9 |
+
|
| 10 |
+
' -%}
|
| 11 |
+
{%- endif -%}
|
| 12 |
+
{%- set loop_messages = messages[1:] -%}
|
| 13 |
+
{%- else -%}
|
| 14 |
+
{%- set first_user_prefix = "" -%}
|
| 15 |
+
{%- set loop_messages = messages -%}
|
| 16 |
+
{%- endif -%}
|
| 17 |
+
{%- for message in loop_messages -%}
|
| 18 |
+
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
|
| 19 |
+
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if (message['role'] == 'assistant') -%}
|
| 22 |
+
{%- set role = "model" -%}
|
| 23 |
+
{%- else -%}
|
| 24 |
+
{%- set role = message['role'] -%}
|
| 25 |
+
{%- endif -%}
|
| 26 |
+
{{ '<start_of_turn>' + role + '
|
| 27 |
+
' + (first_user_prefix if loop.first else "") }}
|
| 28 |
+
{%- if message['content'] is string -%}
|
| 29 |
+
{{ message['content'] | trim }}
|
| 30 |
+
{%- elif message['content'] is iterable -%}
|
| 31 |
+
{%- for item in message['content'] -%}
|
| 32 |
+
{%- if item['type'] == 'image' -%}
|
| 33 |
+
{{ '<start_of_image>' }}
|
| 34 |
+
{%- elif item['type'] == 'text' -%}
|
| 35 |
+
{{ item['text'] | trim }}
|
| 36 |
+
{%- endif -%}
|
| 37 |
+
{%- endfor -%}
|
| 38 |
+
{%- else -%}
|
| 39 |
+
{{ raise_exception("Invalid content type") }}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{ '<end_of_turn>
|
| 42 |
+
' }}
|
| 43 |
+
{%- endfor -%}
|
| 44 |
+
{%- if add_generation_prompt -%}
|
| 45 |
+
{{'<start_of_turn>model
|
| 46 |
+
'}}
|
| 47 |
+
{%- endif -%}
|
full_aft_midtrain4/balanced/checkpoint-32/config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Gemma3ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 255999,
|
| 6 |
+
"bos_token_id": 2,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eoi_token_index": 256000,
|
| 9 |
+
"eos_token_id": 1,
|
| 10 |
+
"image_token_index": 262144,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"mm_tokens_per_image": 256,
|
| 13 |
+
"model_type": "gemma3",
|
| 14 |
+
"pad_token_id": 0,
|
| 15 |
+
"text_config": {
|
| 16 |
+
"_sliding_window_pattern": 6,
|
| 17 |
+
"attention_bias": false,
|
| 18 |
+
"attention_dropout": 0.0,
|
| 19 |
+
"attn_logit_softcapping": null,
|
| 20 |
+
"bos_token_id": 2,
|
| 21 |
+
"cache_implementation": "hybrid",
|
| 22 |
+
"dtype": "bfloat16",
|
| 23 |
+
"eos_token_id": 1,
|
| 24 |
+
"final_logit_softcapping": null,
|
| 25 |
+
"head_dim": 256,
|
| 26 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 27 |
+
"hidden_size": 3840,
|
| 28 |
+
"initializer_range": 0.02,
|
| 29 |
+
"intermediate_size": 15360,
|
| 30 |
+
"layer_types": [
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"sliding_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"sliding_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"sliding_attention",
|
| 59 |
+
"sliding_attention",
|
| 60 |
+
"full_attention",
|
| 61 |
+
"sliding_attention",
|
| 62 |
+
"sliding_attention",
|
| 63 |
+
"sliding_attention",
|
| 64 |
+
"sliding_attention",
|
| 65 |
+
"sliding_attention",
|
| 66 |
+
"full_attention",
|
| 67 |
+
"sliding_attention",
|
| 68 |
+
"sliding_attention",
|
| 69 |
+
"sliding_attention",
|
| 70 |
+
"sliding_attention",
|
| 71 |
+
"sliding_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"sliding_attention",
|
| 74 |
+
"sliding_attention",
|
| 75 |
+
"sliding_attention",
|
| 76 |
+
"sliding_attention",
|
| 77 |
+
"sliding_attention",
|
| 78 |
+
"full_attention"
|
| 79 |
+
],
|
| 80 |
+
"max_position_embeddings": 131072,
|
| 81 |
+
"model_type": "gemma3_text",
|
| 82 |
+
"num_attention_heads": 16,
|
| 83 |
+
"num_hidden_layers": 48,
|
| 84 |
+
"num_key_value_heads": 8,
|
| 85 |
+
"pad_token_id": 0,
|
| 86 |
+
"query_pre_attn_scalar": 256,
|
| 87 |
+
"rms_norm_eps": 1e-06,
|
| 88 |
+
"rope_parameters": {
|
| 89 |
+
"full_attention": {
|
| 90 |
+
"factor": 8.0,
|
| 91 |
+
"rope_theta": 1000000.0,
|
| 92 |
+
"rope_type": "linear"
|
| 93 |
+
},
|
| 94 |
+
"sliding_attention": {
|
| 95 |
+
"rope_theta": 10000.0,
|
| 96 |
+
"rope_type": "default"
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"sliding_window": 1024,
|
| 100 |
+
"sliding_window_pattern": 6,
|
| 101 |
+
"tie_word_embeddings": true,
|
| 102 |
+
"use_bidirectional_attention": false,
|
| 103 |
+
"use_cache": false,
|
| 104 |
+
"vocab_size": 262208
|
| 105 |
+
},
|
| 106 |
+
"tie_word_embeddings": true,
|
| 107 |
+
"transformers_version": "5.9.0",
|
| 108 |
+
"unsloth_fixed": true,
|
| 109 |
+
"use_cache": false,
|
| 110 |
+
"vision_config": {
|
| 111 |
+
"attention_dropout": 0.0,
|
| 112 |
+
"dtype": "bfloat16",
|
| 113 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 114 |
+
"hidden_size": 1152,
|
| 115 |
+
"image_size": 896,
|
| 116 |
+
"intermediate_size": 4304,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"model_type": "siglip_vision_model",
|
| 119 |
+
"num_attention_heads": 16,
|
| 120 |
+
"num_channels": 3,
|
| 121 |
+
"num_hidden_layers": 27,
|
| 122 |
+
"patch_size": 14,
|
| 123 |
+
"vision_use_head": false
|
| 124 |
+
}
|
| 125 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 2,
|
| 3 |
+
"cache_implementation": "hybrid",
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
1,
|
| 7 |
+
106
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 0,
|
| 10 |
+
"top_k": 64,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.9.0"
|
| 13 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71c03dd00e78c43dc4052cc08a04dd8f58588503ee28dd0f7ba7476925c61d4d
|
| 3 |
+
size 26388552360
|
full_aft_midtrain4/balanced/checkpoint-32/preprocessor_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pan_and_scan": null,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": true,
|
| 7 |
+
"image_mean": [
|
| 8 |
+
0.5,
|
| 9 |
+
0.5,
|
| 10 |
+
0.5
|
| 11 |
+
],
|
| 12 |
+
"image_processor_type": "Gemma3ImageProcessor",
|
| 13 |
+
"image_seq_length": 256,
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.5,
|
| 16 |
+
0.5,
|
| 17 |
+
0.5
|
| 18 |
+
],
|
| 19 |
+
"pan_and_scan_max_num_crops": null,
|
| 20 |
+
"pan_and_scan_min_crop_size": null,
|
| 21 |
+
"pan_and_scan_min_ratio_to_activate": null,
|
| 22 |
+
"processor_class": "Gemma3Processor",
|
| 23 |
+
"resample": 2,
|
| 24 |
+
"rescale_factor": 0.00392156862745098,
|
| 25 |
+
"size": {
|
| 26 |
+
"height": 896,
|
| 27 |
+
"width": 896
|
| 28 |
+
}
|
| 29 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/processor_config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_seq_length": 256,
|
| 3 |
+
"processor_class": "Gemma3Processor"
|
| 4 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
|
| 3 |
+
size 33384567
|
full_aft_midtrain4/balanced/checkpoint-32/tokenizer_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"boi_token": "<start_of_image>",
|
| 4 |
+
"bos_token": "<bos>",
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eoi_token": "<end_of_image>",
|
| 7 |
+
"eos_token": "<eos>",
|
| 8 |
+
"image_token": "<image_soft_token>",
|
| 9 |
+
"is_local": false,
|
| 10 |
+
"local_files_only": false,
|
| 11 |
+
"mask_token": "<mask>",
|
| 12 |
+
"model_max_length": 131072,
|
| 13 |
+
"model_specific_special_tokens": {
|
| 14 |
+
"boi_token": "<start_of_image>",
|
| 15 |
+
"eoi_token": "<end_of_image>",
|
| 16 |
+
"image_token": "<image_soft_token>"
|
| 17 |
+
},
|
| 18 |
+
"pad_token": "<pad>",
|
| 19 |
+
"padding_side": "left",
|
| 20 |
+
"processor_class": "Gemma3Processor",
|
| 21 |
+
"sp_model_kwargs": null,
|
| 22 |
+
"spaces_between_special_tokens": false,
|
| 23 |
+
"tokenizer_class": "GemmaTokenizer",
|
| 24 |
+
"unk_token": "<unk>",
|
| 25 |
+
"use_default_system_prompt": false
|
| 26 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/tokens_state.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"total": 696525, "trainable": 14565}
|
full_aft_midtrain4/balanced/checkpoint-32/trainer_state.json
ADDED
|
@@ -0,0 +1,482 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.125,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 32,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.00390625,
|
| 14 |
+
"grad_norm": 17.375,
|
| 15 |
+
"learning_rate": 5e-06,
|
| 16 |
+
"loss": 0.11627960205078125,
|
| 17 |
+
"memory/device_reserved (GiB)": 46.13,
|
| 18 |
+
"memory/max_active (GiB)": 35.94,
|
| 19 |
+
"memory/max_allocated (GiB)": 35.94,
|
| 20 |
+
"ppl": 1.12331,
|
| 21 |
+
"step": 1,
|
| 22 |
+
"tokens/total": 22614,
|
| 23 |
+
"tokens/train_per_sec_per_gpu": 1.99,
|
| 24 |
+
"tokens/trainable": 470
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.0078125,
|
| 28 |
+
"grad_norm": 1472.0,
|
| 29 |
+
"learning_rate": 5e-06,
|
| 30 |
+
"loss": 0.1474212110042572,
|
| 31 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 32 |
+
"memory/max_active (GiB)": 46.9,
|
| 33 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 34 |
+
"ppl": 1.15884,
|
| 35 |
+
"step": 2,
|
| 36 |
+
"tokens/total": 44116,
|
| 37 |
+
"tokens/train_per_sec_per_gpu": 4.14,
|
| 38 |
+
"tokens/trainable": 922
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.01171875,
|
| 42 |
+
"grad_norm": 57.75,
|
| 43 |
+
"learning_rate": 5e-06,
|
| 44 |
+
"loss": 0.2154746651649475,
|
| 45 |
+
"memory/device_reserved (GiB)": 58.08,
|
| 46 |
+
"memory/max_active (GiB)": 46.9,
|
| 47 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 48 |
+
"ppl": 1.24045,
|
| 49 |
+
"step": 3,
|
| 50 |
+
"tokens/total": 67092,
|
| 51 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 52 |
+
"tokens/trainable": 1396
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"epoch": 0.015625,
|
| 56 |
+
"grad_norm": 37.75,
|
| 57 |
+
"learning_rate": 5e-06,
|
| 58 |
+
"loss": 0.10307750105857849,
|
| 59 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 60 |
+
"memory/max_active (GiB)": 46.9,
|
| 61 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 62 |
+
"ppl": 1.10858,
|
| 63 |
+
"step": 4,
|
| 64 |
+
"tokens/total": 88800,
|
| 65 |
+
"tokens/train_per_sec_per_gpu": 5.16,
|
| 66 |
+
"tokens/trainable": 1850
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"epoch": 0.01953125,
|
| 70 |
+
"grad_norm": 25.0,
|
| 71 |
+
"learning_rate": 5e-06,
|
| 72 |
+
"loss": 0.10710409283638,
|
| 73 |
+
"memory/device_reserved (GiB)": 58.09,
|
| 74 |
+
"memory/max_active (GiB)": 46.9,
|
| 75 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 76 |
+
"ppl": 1.11305,
|
| 77 |
+
"step": 5,
|
| 78 |
+
"tokens/total": 110742,
|
| 79 |
+
"tokens/train_per_sec_per_gpu": 3.92,
|
| 80 |
+
"tokens/trainable": 2302
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.0234375,
|
| 84 |
+
"grad_norm": 16.25,
|
| 85 |
+
"learning_rate": 5e-06,
|
| 86 |
+
"loss": 0.05220422148704529,
|
| 87 |
+
"memory/device_reserved (GiB)": 57.19,
|
| 88 |
+
"memory/max_active (GiB)": 46.9,
|
| 89 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 90 |
+
"ppl": 1.05359,
|
| 91 |
+
"step": 6,
|
| 92 |
+
"tokens/total": 131497,
|
| 93 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 94 |
+
"tokens/trainable": 2742
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"epoch": 0.02734375,
|
| 98 |
+
"grad_norm": 18.0,
|
| 99 |
+
"learning_rate": 5e-06,
|
| 100 |
+
"loss": 0.06188321113586426,
|
| 101 |
+
"memory/device_reserved (GiB)": 59.04,
|
| 102 |
+
"memory/max_active (GiB)": 46.9,
|
| 103 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 104 |
+
"ppl": 1.06384,
|
| 105 |
+
"step": 7,
|
| 106 |
+
"tokens/total": 153846,
|
| 107 |
+
"tokens/train_per_sec_per_gpu": 4.99,
|
| 108 |
+
"tokens/trainable": 3208
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 0.03125,
|
| 112 |
+
"grad_norm": 16.75,
|
| 113 |
+
"learning_rate": 5e-06,
|
| 114 |
+
"loss": 0.036078453063964844,
|
| 115 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 116 |
+
"memory/max_active (GiB)": 46.9,
|
| 117 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 118 |
+
"ppl": 1.03674,
|
| 119 |
+
"step": 8,
|
| 120 |
+
"tokens/total": 175713,
|
| 121 |
+
"tokens/train_per_sec_per_gpu": 4.47,
|
| 122 |
+
"tokens/trainable": 3655
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.03515625,
|
| 126 |
+
"grad_norm": 13.5625,
|
| 127 |
+
"learning_rate": 5e-06,
|
| 128 |
+
"loss": 0.05783689022064209,
|
| 129 |
+
"memory/device_reserved (GiB)": 60.12,
|
| 130 |
+
"memory/max_active (GiB)": 46.9,
|
| 131 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 132 |
+
"ppl": 1.05954,
|
| 133 |
+
"step": 9,
|
| 134 |
+
"tokens/total": 196358,
|
| 135 |
+
"tokens/train_per_sec_per_gpu": 3.39,
|
| 136 |
+
"tokens/trainable": 4091
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"epoch": 0.0390625,
|
| 140 |
+
"grad_norm": 16.375,
|
| 141 |
+
"learning_rate": 5e-06,
|
| 142 |
+
"loss": 0.05202341079711914,
|
| 143 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 144 |
+
"memory/max_active (GiB)": 46.9,
|
| 145 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 146 |
+
"ppl": 1.0534,
|
| 147 |
+
"step": 10,
|
| 148 |
+
"tokens/total": 218622,
|
| 149 |
+
"tokens/train_per_sec_per_gpu": 4.33,
|
| 150 |
+
"tokens/trainable": 4553
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 0.04296875,
|
| 154 |
+
"grad_norm": 22.5,
|
| 155 |
+
"learning_rate": 5e-06,
|
| 156 |
+
"loss": 0.03920435905456543,
|
| 157 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 158 |
+
"memory/max_active (GiB)": 46.9,
|
| 159 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 160 |
+
"ppl": 1.03998,
|
| 161 |
+
"step": 11,
|
| 162 |
+
"tokens/total": 239835,
|
| 163 |
+
"tokens/train_per_sec_per_gpu": 4.54,
|
| 164 |
+
"tokens/trainable": 4992
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.046875,
|
| 168 |
+
"grad_norm": 43.75,
|
| 169 |
+
"learning_rate": 5e-06,
|
| 170 |
+
"loss": 0.0486750602722168,
|
| 171 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 172 |
+
"memory/max_active (GiB)": 46.9,
|
| 173 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 174 |
+
"ppl": 1.04988,
|
| 175 |
+
"step": 12,
|
| 176 |
+
"tokens/total": 264428,
|
| 177 |
+
"tokens/train_per_sec_per_gpu": 4.93,
|
| 178 |
+
"tokens/trainable": 5494
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"epoch": 0.05078125,
|
| 182 |
+
"grad_norm": 7.59375,
|
| 183 |
+
"learning_rate": 5e-06,
|
| 184 |
+
"loss": 0.023757457733154297,
|
| 185 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 186 |
+
"memory/max_active (GiB)": 46.9,
|
| 187 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 188 |
+
"ppl": 1.02404,
|
| 189 |
+
"step": 13,
|
| 190 |
+
"tokens/total": 284859,
|
| 191 |
+
"tokens/train_per_sec_per_gpu": 3.95,
|
| 192 |
+
"tokens/trainable": 5933
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"epoch": 0.0546875,
|
| 196 |
+
"grad_norm": 19.0,
|
| 197 |
+
"learning_rate": 5e-06,
|
| 198 |
+
"loss": 0.04330146312713623,
|
| 199 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 200 |
+
"memory/max_active (GiB)": 46.9,
|
| 201 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 202 |
+
"ppl": 1.04425,
|
| 203 |
+
"step": 14,
|
| 204 |
+
"tokens/total": 305640,
|
| 205 |
+
"tokens/train_per_sec_per_gpu": 3.17,
|
| 206 |
+
"tokens/trainable": 6369
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.05859375,
|
| 210 |
+
"grad_norm": 9.6875,
|
| 211 |
+
"learning_rate": 5e-06,
|
| 212 |
+
"loss": 0.03816962242126465,
|
| 213 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 214 |
+
"memory/max_active (GiB)": 46.9,
|
| 215 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 216 |
+
"ppl": 1.03891,
|
| 217 |
+
"step": 15,
|
| 218 |
+
"tokens/total": 327063,
|
| 219 |
+
"tokens/train_per_sec_per_gpu": 4.35,
|
| 220 |
+
"tokens/trainable": 6820
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.0625,
|
| 224 |
+
"grad_norm": 11.8125,
|
| 225 |
+
"learning_rate": 5e-06,
|
| 226 |
+
"loss": 0.03375864028930664,
|
| 227 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 228 |
+
"memory/max_active (GiB)": 46.9,
|
| 229 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 230 |
+
"ppl": 1.03433,
|
| 231 |
+
"step": 16,
|
| 232 |
+
"tokens/total": 347838,
|
| 233 |
+
"tokens/train_per_sec_per_gpu": 3.77,
|
| 234 |
+
"tokens/trainable": 7258
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.06640625,
|
| 238 |
+
"grad_norm": 12.4375,
|
| 239 |
+
"learning_rate": 5e-06,
|
| 240 |
+
"loss": 0.023131251335144043,
|
| 241 |
+
"memory/device_reserved (GiB)": 60.1,
|
| 242 |
+
"memory/max_active (GiB)": 46.9,
|
| 243 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 244 |
+
"ppl": 1.0234,
|
| 245 |
+
"step": 17,
|
| 246 |
+
"tokens/total": 369888,
|
| 247 |
+
"tokens/train_per_sec_per_gpu": 4.4,
|
| 248 |
+
"tokens/trainable": 7710
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 0.0703125,
|
| 252 |
+
"grad_norm": 10.75,
|
| 253 |
+
"learning_rate": 5e-06,
|
| 254 |
+
"loss": 0.02522754669189453,
|
| 255 |
+
"memory/device_reserved (GiB)": 57.25,
|
| 256 |
+
"memory/max_active (GiB)": 46.9,
|
| 257 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 258 |
+
"ppl": 1.02555,
|
| 259 |
+
"step": 18,
|
| 260 |
+
"tokens/total": 392455,
|
| 261 |
+
"tokens/train_per_sec_per_gpu": 3.72,
|
| 262 |
+
"tokens/trainable": 8174
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.07421875,
|
| 266 |
+
"grad_norm": 8.4375,
|
| 267 |
+
"learning_rate": 5e-06,
|
| 268 |
+
"loss": 0.028311967849731445,
|
| 269 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 270 |
+
"memory/max_active (GiB)": 46.9,
|
| 271 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 272 |
+
"ppl": 1.02872,
|
| 273 |
+
"step": 19,
|
| 274 |
+
"tokens/total": 413549,
|
| 275 |
+
"tokens/train_per_sec_per_gpu": 4.51,
|
| 276 |
+
"tokens/trainable": 8617
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"epoch": 0.078125,
|
| 280 |
+
"grad_norm": 4.34375,
|
| 281 |
+
"learning_rate": 5e-06,
|
| 282 |
+
"loss": 0.012633323669433594,
|
| 283 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 284 |
+
"memory/max_active (GiB)": 46.9,
|
| 285 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 286 |
+
"ppl": 1.01271,
|
| 287 |
+
"step": 20,
|
| 288 |
+
"tokens/total": 434128,
|
| 289 |
+
"tokens/train_per_sec_per_gpu": 3.7,
|
| 290 |
+
"tokens/trainable": 9037
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 0.08203125,
|
| 294 |
+
"grad_norm": 3.34375,
|
| 295 |
+
"learning_rate": 5e-06,
|
| 296 |
+
"loss": 0.010331928730010986,
|
| 297 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 298 |
+
"memory/max_active (GiB)": 46.9,
|
| 299 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 300 |
+
"ppl": 1.01039,
|
| 301 |
+
"step": 21,
|
| 302 |
+
"tokens/total": 456131,
|
| 303 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 304 |
+
"tokens/trainable": 9503
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"epoch": 0.0859375,
|
| 308 |
+
"grad_norm": 8.6875,
|
| 309 |
+
"learning_rate": 5e-06,
|
| 310 |
+
"loss": 0.011040687561035156,
|
| 311 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 312 |
+
"memory/max_active (GiB)": 46.9,
|
| 313 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 314 |
+
"ppl": 1.0111,
|
| 315 |
+
"step": 22,
|
| 316 |
+
"tokens/total": 478707,
|
| 317 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 318 |
+
"tokens/trainable": 9972
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 0.08984375,
|
| 322 |
+
"grad_norm": 17.75,
|
| 323 |
+
"learning_rate": 5e-06,
|
| 324 |
+
"loss": 0.030008912086486816,
|
| 325 |
+
"memory/device_reserved (GiB)": 58.4,
|
| 326 |
+
"memory/max_active (GiB)": 46.9,
|
| 327 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 328 |
+
"ppl": 1.03046,
|
| 329 |
+
"step": 23,
|
| 330 |
+
"tokens/total": 501425,
|
| 331 |
+
"tokens/train_per_sec_per_gpu": 4.9,
|
| 332 |
+
"tokens/trainable": 10447
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 0.09375,
|
| 336 |
+
"grad_norm": 6.6875,
|
| 337 |
+
"learning_rate": 5e-06,
|
| 338 |
+
"loss": 0.007420599460601807,
|
| 339 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 340 |
+
"memory/max_active (GiB)": 46.9,
|
| 341 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 342 |
+
"ppl": 1.00745,
|
| 343 |
+
"step": 24,
|
| 344 |
+
"tokens/total": 522885,
|
| 345 |
+
"tokens/train_per_sec_per_gpu": 5.15,
|
| 346 |
+
"tokens/trainable": 10899
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 0.09765625,
|
| 350 |
+
"grad_norm": 24.75,
|
| 351 |
+
"learning_rate": 5e-06,
|
| 352 |
+
"loss": 0.05265447497367859,
|
| 353 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 354 |
+
"memory/max_active (GiB)": 46.9,
|
| 355 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 356 |
+
"ppl": 1.05407,
|
| 357 |
+
"step": 25,
|
| 358 |
+
"tokens/total": 544886,
|
| 359 |
+
"tokens/train_per_sec_per_gpu": 4.45,
|
| 360 |
+
"tokens/trainable": 11360
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"epoch": 0.1015625,
|
| 364 |
+
"grad_norm": 12.625,
|
| 365 |
+
"learning_rate": 5e-06,
|
| 366 |
+
"loss": 0.04731038212776184,
|
| 367 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 368 |
+
"memory/max_active (GiB)": 46.9,
|
| 369 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 370 |
+
"ppl": 1.04845,
|
| 371 |
+
"step": 26,
|
| 372 |
+
"tokens/total": 564964,
|
| 373 |
+
"tokens/train_per_sec_per_gpu": 3.36,
|
| 374 |
+
"tokens/trainable": 11775
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 0.10546875,
|
| 378 |
+
"grad_norm": 13.1875,
|
| 379 |
+
"learning_rate": 5e-06,
|
| 380 |
+
"loss": 0.014921486377716064,
|
| 381 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 382 |
+
"memory/max_active (GiB)": 46.9,
|
| 383 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 384 |
+
"ppl": 1.01503,
|
| 385 |
+
"step": 27,
|
| 386 |
+
"tokens/total": 586567,
|
| 387 |
+
"tokens/train_per_sec_per_gpu": 5.29,
|
| 388 |
+
"tokens/trainable": 12242
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.109375,
|
| 392 |
+
"grad_norm": 18.625,
|
| 393 |
+
"learning_rate": 5e-06,
|
| 394 |
+
"loss": 0.013006031513214111,
|
| 395 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 396 |
+
"memory/max_active (GiB)": 46.9,
|
| 397 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 398 |
+
"ppl": 1.01309,
|
| 399 |
+
"step": 28,
|
| 400 |
+
"tokens/total": 608188,
|
| 401 |
+
"tokens/train_per_sec_per_gpu": 3.78,
|
| 402 |
+
"tokens/trainable": 12696
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 0.11328125,
|
| 406 |
+
"grad_norm": 21.25,
|
| 407 |
+
"learning_rate": 5e-06,
|
| 408 |
+
"loss": 0.028101369738578796,
|
| 409 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 410 |
+
"memory/max_active (GiB)": 46.9,
|
| 411 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 412 |
+
"ppl": 1.0285,
|
| 413 |
+
"step": 29,
|
| 414 |
+
"tokens/total": 630007,
|
| 415 |
+
"tokens/train_per_sec_per_gpu": 3.81,
|
| 416 |
+
"tokens/trainable": 13153
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.1171875,
|
| 420 |
+
"grad_norm": 8.3125,
|
| 421 |
+
"learning_rate": 5e-06,
|
| 422 |
+
"loss": 0.01690027117729187,
|
| 423 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 424 |
+
"memory/max_active (GiB)": 46.9,
|
| 425 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 426 |
+
"ppl": 1.01704,
|
| 427 |
+
"step": 30,
|
| 428 |
+
"tokens/total": 651656,
|
| 429 |
+
"tokens/train_per_sec_per_gpu": 5.16,
|
| 430 |
+
"tokens/trainable": 13624
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"epoch": 0.12109375,
|
| 434 |
+
"grad_norm": 12.3125,
|
| 435 |
+
"learning_rate": 5e-06,
|
| 436 |
+
"loss": 0.015246853232383728,
|
| 437 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 438 |
+
"memory/max_active (GiB)": 46.9,
|
| 439 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 440 |
+
"ppl": 1.01536,
|
| 441 |
+
"step": 31,
|
| 442 |
+
"tokens/total": 672938,
|
| 443 |
+
"tokens/train_per_sec_per_gpu": 5.14,
|
| 444 |
+
"tokens/trainable": 14064
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.125,
|
| 448 |
+
"grad_norm": 11.125,
|
| 449 |
+
"learning_rate": 5e-06,
|
| 450 |
+
"loss": 0.017684906721115112,
|
| 451 |
+
"memory/device_reserved (GiB)": 60.14,
|
| 452 |
+
"memory/max_active (GiB)": 46.9,
|
| 453 |
+
"memory/max_allocated (GiB)": 46.9,
|
| 454 |
+
"ppl": 1.01784,
|
| 455 |
+
"step": 32,
|
| 456 |
+
"tokens/total": 696525,
|
| 457 |
+
"tokens/train_per_sec_per_gpu": 5.03,
|
| 458 |
+
"tokens/trainable": 14565
|
| 459 |
+
}
|
| 460 |
+
],
|
| 461 |
+
"logging_steps": 1,
|
| 462 |
+
"max_steps": 512,
|
| 463 |
+
"num_input_tokens_seen": 0,
|
| 464 |
+
"num_train_epochs": 2,
|
| 465 |
+
"save_steps": 500,
|
| 466 |
+
"stateful_callbacks": {
|
| 467 |
+
"TrainerControl": {
|
| 468 |
+
"args": {
|
| 469 |
+
"should_epoch_stop": false,
|
| 470 |
+
"should_evaluate": false,
|
| 471 |
+
"should_log": false,
|
| 472 |
+
"should_save": true,
|
| 473 |
+
"should_training_stop": false
|
| 474 |
+
},
|
| 475 |
+
"attributes": {}
|
| 476 |
+
}
|
| 477 |
+
},
|
| 478 |
+
"total_flos": 4.670504244242022e+16,
|
| 479 |
+
"train_batch_size": 1,
|
| 480 |
+
"trial_name": null,
|
| 481 |
+
"trial_params": null
|
| 482 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-32/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0107cd88280e968b63f91f7125176eba0c9002c1f534529b1c400b807d0860f
|
| 3 |
+
size 8913
|
full_aft_midtrain4/balanced/checkpoint-4/chat_template.jinja
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{ bos_token }}
|
| 2 |
+
{%- if messages[0]['role'] == 'system' -%}
|
| 3 |
+
{%- if messages[0]['content'] is string -%}
|
| 4 |
+
{%- set first_user_prefix = messages[0]['content'] + '
|
| 5 |
+
|
| 6 |
+
' -%}
|
| 7 |
+
{%- else -%}
|
| 8 |
+
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
|
| 9 |
+
|
| 10 |
+
' -%}
|
| 11 |
+
{%- endif -%}
|
| 12 |
+
{%- set loop_messages = messages[1:] -%}
|
| 13 |
+
{%- else -%}
|
| 14 |
+
{%- set first_user_prefix = "" -%}
|
| 15 |
+
{%- set loop_messages = messages -%}
|
| 16 |
+
{%- endif -%}
|
| 17 |
+
{%- for message in loop_messages -%}
|
| 18 |
+
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
|
| 19 |
+
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if (message['role'] == 'assistant') -%}
|
| 22 |
+
{%- set role = "model" -%}
|
| 23 |
+
{%- else -%}
|
| 24 |
+
{%- set role = message['role'] -%}
|
| 25 |
+
{%- endif -%}
|
| 26 |
+
{{ '<start_of_turn>' + role + '
|
| 27 |
+
' + (first_user_prefix if loop.first else "") }}
|
| 28 |
+
{%- if message['content'] is string -%}
|
| 29 |
+
{{ message['content'] | trim }}
|
| 30 |
+
{%- elif message['content'] is iterable -%}
|
| 31 |
+
{%- for item in message['content'] -%}
|
| 32 |
+
{%- if item['type'] == 'image' -%}
|
| 33 |
+
{{ '<start_of_image>' }}
|
| 34 |
+
{%- elif item['type'] == 'text' -%}
|
| 35 |
+
{{ item['text'] | trim }}
|
| 36 |
+
{%- endif -%}
|
| 37 |
+
{%- endfor -%}
|
| 38 |
+
{%- else -%}
|
| 39 |
+
{{ raise_exception("Invalid content type") }}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{ '<end_of_turn>
|
| 42 |
+
' }}
|
| 43 |
+
{%- endfor -%}
|
| 44 |
+
{%- if add_generation_prompt -%}
|
| 45 |
+
{{'<start_of_turn>model
|
| 46 |
+
'}}
|
| 47 |
+
{%- endif -%}
|
full_aft_midtrain4/balanced/checkpoint-4/config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Gemma3ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 255999,
|
| 6 |
+
"bos_token_id": 2,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eoi_token_index": 256000,
|
| 9 |
+
"eos_token_id": 1,
|
| 10 |
+
"image_token_index": 262144,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"mm_tokens_per_image": 256,
|
| 13 |
+
"model_type": "gemma3",
|
| 14 |
+
"pad_token_id": 0,
|
| 15 |
+
"text_config": {
|
| 16 |
+
"_sliding_window_pattern": 6,
|
| 17 |
+
"attention_bias": false,
|
| 18 |
+
"attention_dropout": 0.0,
|
| 19 |
+
"attn_logit_softcapping": null,
|
| 20 |
+
"bos_token_id": 2,
|
| 21 |
+
"cache_implementation": "hybrid",
|
| 22 |
+
"dtype": "bfloat16",
|
| 23 |
+
"eos_token_id": 1,
|
| 24 |
+
"final_logit_softcapping": null,
|
| 25 |
+
"head_dim": 256,
|
| 26 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 27 |
+
"hidden_size": 3840,
|
| 28 |
+
"initializer_range": 0.02,
|
| 29 |
+
"intermediate_size": 15360,
|
| 30 |
+
"layer_types": [
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"sliding_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"sliding_attention",
|
| 40 |
+
"sliding_attention",
|
| 41 |
+
"sliding_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"sliding_attention",
|
| 44 |
+
"sliding_attention",
|
| 45 |
+
"sliding_attention",
|
| 46 |
+
"sliding_attention",
|
| 47 |
+
"sliding_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"sliding_attention",
|
| 50 |
+
"sliding_attention",
|
| 51 |
+
"sliding_attention",
|
| 52 |
+
"sliding_attention",
|
| 53 |
+
"sliding_attention",
|
| 54 |
+
"full_attention",
|
| 55 |
+
"sliding_attention",
|
| 56 |
+
"sliding_attention",
|
| 57 |
+
"sliding_attention",
|
| 58 |
+
"sliding_attention",
|
| 59 |
+
"sliding_attention",
|
| 60 |
+
"full_attention",
|
| 61 |
+
"sliding_attention",
|
| 62 |
+
"sliding_attention",
|
| 63 |
+
"sliding_attention",
|
| 64 |
+
"sliding_attention",
|
| 65 |
+
"sliding_attention",
|
| 66 |
+
"full_attention",
|
| 67 |
+
"sliding_attention",
|
| 68 |
+
"sliding_attention",
|
| 69 |
+
"sliding_attention",
|
| 70 |
+
"sliding_attention",
|
| 71 |
+
"sliding_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"sliding_attention",
|
| 74 |
+
"sliding_attention",
|
| 75 |
+
"sliding_attention",
|
| 76 |
+
"sliding_attention",
|
| 77 |
+
"sliding_attention",
|
| 78 |
+
"full_attention"
|
| 79 |
+
],
|
| 80 |
+
"max_position_embeddings": 131072,
|
| 81 |
+
"model_type": "gemma3_text",
|
| 82 |
+
"num_attention_heads": 16,
|
| 83 |
+
"num_hidden_layers": 48,
|
| 84 |
+
"num_key_value_heads": 8,
|
| 85 |
+
"pad_token_id": 0,
|
| 86 |
+
"query_pre_attn_scalar": 256,
|
| 87 |
+
"rms_norm_eps": 1e-06,
|
| 88 |
+
"rope_parameters": {
|
| 89 |
+
"full_attention": {
|
| 90 |
+
"factor": 8.0,
|
| 91 |
+
"rope_theta": 1000000.0,
|
| 92 |
+
"rope_type": "linear"
|
| 93 |
+
},
|
| 94 |
+
"sliding_attention": {
|
| 95 |
+
"rope_theta": 10000.0,
|
| 96 |
+
"rope_type": "default"
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"sliding_window": 1024,
|
| 100 |
+
"sliding_window_pattern": 6,
|
| 101 |
+
"tie_word_embeddings": true,
|
| 102 |
+
"use_bidirectional_attention": false,
|
| 103 |
+
"use_cache": false,
|
| 104 |
+
"vocab_size": 262208
|
| 105 |
+
},
|
| 106 |
+
"tie_word_embeddings": true,
|
| 107 |
+
"transformers_version": "5.9.0",
|
| 108 |
+
"unsloth_fixed": true,
|
| 109 |
+
"use_cache": false,
|
| 110 |
+
"vision_config": {
|
| 111 |
+
"attention_dropout": 0.0,
|
| 112 |
+
"dtype": "bfloat16",
|
| 113 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 114 |
+
"hidden_size": 1152,
|
| 115 |
+
"image_size": 896,
|
| 116 |
+
"intermediate_size": 4304,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"model_type": "siglip_vision_model",
|
| 119 |
+
"num_attention_heads": 16,
|
| 120 |
+
"num_channels": 3,
|
| 121 |
+
"num_hidden_layers": 27,
|
| 122 |
+
"patch_size": 14,
|
| 123 |
+
"vision_use_head": false
|
| 124 |
+
}
|
| 125 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-4/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 2,
|
| 3 |
+
"cache_implementation": "hybrid",
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
1,
|
| 7 |
+
106
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 0,
|
| 10 |
+
"top_k": 64,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.9.0"
|
| 13 |
+
}
|
full_aft_midtrain4/balanced/checkpoint-4/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d4a0f1216763f701d3156eab5d12dd3f2ee3a75d28d01e2e80373cca8e39625
|
| 3 |
+
size 26388552360
|
full_aft_midtrain4/balanced/checkpoint-4/preprocessor_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pan_and_scan": null,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": true,
|
| 7 |
+
"image_mean": [
|
| 8 |
+
0.5,
|
| 9 |
+
0.5,
|
| 10 |
+
0.5
|
| 11 |
+
],
|
| 12 |
+
"image_processor_type": "Gemma3ImageProcessor",
|
| 13 |
+
"image_seq_length": 256,
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.5,
|
| 16 |
+
0.5,
|
| 17 |
+
0.5
|
| 18 |
+
],
|
| 19 |
+
"pan_and_scan_max_num_crops": null,
|
| 20 |
+
"pan_and_scan_min_crop_size": null,
|
| 21 |
+
"pan_and_scan_min_ratio_to_activate": null,
|
| 22 |
+
"processor_class": "Gemma3Processor",
|
| 23 |
+
"resample": 2,
|
| 24 |
+
"rescale_factor": 0.00392156862745098,
|
| 25 |
+
"size": {
|
| 26 |
+
"height": 896,
|
| 27 |
+
"width": 896
|
| 28 |
+
}
|
| 29 |
+
}
|