grail_v1_343_b64_sft_e3_ck300
Browse files- .gitattributes +1 -0
- grail_v1_343_b64_sft_e3_ck300/all_results.json +12 -0
- grail_v1_343_b64_sft_e3_ck300/chat_template.jinja +120 -0
- grail_v1_343_b64_sft_e3_ck300/config.json +70 -0
- grail_v1_343_b64_sft_e3_ck300/eval_results.json +7 -0
- grail_v1_343_b64_sft_e3_ck300/generation_config.json +14 -0
- grail_v1_343_b64_sft_e3_ck300/model.safetensors +3 -0
- grail_v1_343_b64_sft_e3_ck300/processor_config.json +60 -0
- grail_v1_343_b64_sft_e3_ck300/sft.yaml +58 -0
- grail_v1_343_b64_sft_e3_ck300/tokenizer.json +3 -0
- grail_v1_343_b64_sft_e3_ck300/tokenizer_config.json +32 -0
- grail_v1_343_b64_sft_e3_ck300/train_results.json +8 -0
- grail_v1_343_b64_sft_e3_ck300/trainer_state.json +277 -0
.gitattributes
CHANGED
|
@@ -72,3 +72,4 @@ her_nemi_graph_343_geo_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge
|
|
| 72 |
her_nemi_graph_343bal_geo_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 73 |
her_nemi_graph_fcur_geo_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 74 |
her_nemi_graph_fcur_rw_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 72 |
her_nemi_graph_343bal_geo_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 73 |
her_nemi_graph_fcur_geo_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 74 |
her_nemi_graph_fcur_rw_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 75 |
+
grail_v1_343_b64_sft_e3_ck300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
grail_v1_343_b64_sft_e3_ck300/all_results.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.1253263707571803,
|
| 3 |
+
"eval_loss": 0.08317559212446213,
|
| 4 |
+
"eval_runtime": 23.7677,
|
| 5 |
+
"eval_samples_per_second": 13.59,
|
| 6 |
+
"eval_steps_per_second": 3.408,
|
| 7 |
+
"total_flos": 138960941613056.0,
|
| 8 |
+
"train_loss": 0.12055180182059606,
|
| 9 |
+
"train_runtime": 3553.0676,
|
| 10 |
+
"train_samples_per_second": 5.404,
|
| 11 |
+
"train_steps_per_second": 0.084
|
| 12 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/chat_template.jinja
ADDED
|
@@ -0,0 +1,120 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- if tools %}
|
| 2 |
+
{{- '<|im_start|>system\n' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{%- if messages[0].content is string %}
|
| 5 |
+
{{- messages[0].content }}
|
| 6 |
+
{%- else %}
|
| 7 |
+
{%- for content in messages[0].content %}
|
| 8 |
+
{%- if 'text' in content %}
|
| 9 |
+
{{- content.text }}
|
| 10 |
+
{%- endif %}
|
| 11 |
+
{%- endfor %}
|
| 12 |
+
{%- endif %}
|
| 13 |
+
{{- '\n\n' }}
|
| 14 |
+
{%- endif %}
|
| 15 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 16 |
+
{%- for tool in tools %}
|
| 17 |
+
{{- "\n" }}
|
| 18 |
+
{{- tool | tojson }}
|
| 19 |
+
{%- endfor %}
|
| 20 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 21 |
+
{%- else %}
|
| 22 |
+
{%- if messages[0].role == 'system' %}
|
| 23 |
+
{{- '<|im_start|>system\n' }}
|
| 24 |
+
{%- if messages[0].content is string %}
|
| 25 |
+
{{- messages[0].content }}
|
| 26 |
+
{%- else %}
|
| 27 |
+
{%- for content in messages[0].content %}
|
| 28 |
+
{%- if 'text' in content %}
|
| 29 |
+
{{- content.text }}
|
| 30 |
+
{%- endif %}
|
| 31 |
+
{%- endfor %}
|
| 32 |
+
{%- endif %}
|
| 33 |
+
{{- '<|im_end|>\n' }}
|
| 34 |
+
{%- endif %}
|
| 35 |
+
{%- endif %}
|
| 36 |
+
{%- set image_count = namespace(value=0) %}
|
| 37 |
+
{%- set video_count = namespace(value=0) %}
|
| 38 |
+
{%- for message in messages %}
|
| 39 |
+
{%- if message.role == "user" %}
|
| 40 |
+
{{- '<|im_start|>' + message.role + '\n' }}
|
| 41 |
+
{%- if message.content is string %}
|
| 42 |
+
{{- message.content }}
|
| 43 |
+
{%- else %}
|
| 44 |
+
{%- for content in message.content %}
|
| 45 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 46 |
+
{%- set image_count.value = image_count.value + 1 %}
|
| 47 |
+
{%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
|
| 48 |
+
<|vision_start|><|image_pad|><|vision_end|>
|
| 49 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 50 |
+
{%- set video_count.value = video_count.value + 1 %}
|
| 51 |
+
{%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
|
| 52 |
+
<|vision_start|><|video_pad|><|vision_end|>
|
| 53 |
+
{%- elif 'text' in content %}
|
| 54 |
+
{{- content.text }}
|
| 55 |
+
{%- endif %}
|
| 56 |
+
{%- endfor %}
|
| 57 |
+
{%- endif %}
|
| 58 |
+
{{- '<|im_end|>\n' }}
|
| 59 |
+
{%- elif message.role == "assistant" %}
|
| 60 |
+
{{- '<|im_start|>' + message.role + '\n' }}
|
| 61 |
+
{%- if message.content is string %}
|
| 62 |
+
{{- message.content }}
|
| 63 |
+
{%- else %}
|
| 64 |
+
{%- for content_item in message.content %}
|
| 65 |
+
{%- if 'text' in content_item %}
|
| 66 |
+
{{- content_item.text }}
|
| 67 |
+
{%- endif %}
|
| 68 |
+
{%- endfor %}
|
| 69 |
+
{%- endif %}
|
| 70 |
+
{%- if message.tool_calls %}
|
| 71 |
+
{%- for tool_call in message.tool_calls %}
|
| 72 |
+
{%- if (loop.first and message.content) or (not loop.first) %}
|
| 73 |
+
{{- '\n' }}
|
| 74 |
+
{%- endif %}
|
| 75 |
+
{%- if tool_call.function %}
|
| 76 |
+
{%- set tool_call = tool_call.function %}
|
| 77 |
+
{%- endif %}
|
| 78 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 79 |
+
{{- tool_call.name }}
|
| 80 |
+
{{- '", "arguments": ' }}
|
| 81 |
+
{%- if tool_call.arguments is string %}
|
| 82 |
+
{{- tool_call.arguments }}
|
| 83 |
+
{%- else %}
|
| 84 |
+
{{- tool_call.arguments | tojson }}
|
| 85 |
+
{%- endif %}
|
| 86 |
+
{{- '}\n</tool_call>' }}
|
| 87 |
+
{%- endfor %}
|
| 88 |
+
{%- endif %}
|
| 89 |
+
{{- '<|im_end|>\n' }}
|
| 90 |
+
{%- elif message.role == "tool" %}
|
| 91 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 92 |
+
{{- '<|im_start|>user' }}
|
| 93 |
+
{%- endif %}
|
| 94 |
+
{{- '\n<tool_response>\n' }}
|
| 95 |
+
{%- if message.content is string %}
|
| 96 |
+
{{- message.content }}
|
| 97 |
+
{%- else %}
|
| 98 |
+
{%- for content in message.content %}
|
| 99 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 100 |
+
{%- set image_count.value = image_count.value + 1 %}
|
| 101 |
+
{%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
|
| 102 |
+
<|vision_start|><|image_pad|><|vision_end|>
|
| 103 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 104 |
+
{%- set video_count.value = video_count.value + 1 %}
|
| 105 |
+
{%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
|
| 106 |
+
<|vision_start|><|video_pad|><|vision_end|>
|
| 107 |
+
{%- elif 'text' in content %}
|
| 108 |
+
{{- content.text }}
|
| 109 |
+
{%- endif %}
|
| 110 |
+
{%- endfor %}
|
| 111 |
+
{%- endif %}
|
| 112 |
+
{{- '\n</tool_response>' }}
|
| 113 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 114 |
+
{{- '<|im_end|>\n' }}
|
| 115 |
+
{%- endif %}
|
| 116 |
+
{%- endif %}
|
| 117 |
+
{%- endfor %}
|
| 118 |
+
{%- if add_generation_prompt %}
|
| 119 |
+
{{- '<|im_start|>assistant\n' }}
|
| 120 |
+
{%- endif %}
|
grail_v1_343_b64_sft_e3_ck300/config.json
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3VLForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"bos_token_id": null,
|
| 6 |
+
"dtype": "bfloat16",
|
| 7 |
+
"eos_token_id": 151645,
|
| 8 |
+
"hidden_size": 4096,
|
| 9 |
+
"image_token_id": 151655,
|
| 10 |
+
"model_type": "qwen3_vl",
|
| 11 |
+
"pad_token_id": 151643,
|
| 12 |
+
"text_config": {
|
| 13 |
+
"attention_bias": false,
|
| 14 |
+
"attention_dropout": 0.0,
|
| 15 |
+
"bos_token_id": 151643,
|
| 16 |
+
"dtype": "bfloat16",
|
| 17 |
+
"eos_token_id": 151645,
|
| 18 |
+
"head_dim": 128,
|
| 19 |
+
"hidden_act": "silu",
|
| 20 |
+
"hidden_size": 4096,
|
| 21 |
+
"initializer_range": 0.02,
|
| 22 |
+
"intermediate_size": 12288,
|
| 23 |
+
"max_position_embeddings": 262144,
|
| 24 |
+
"model_type": "qwen3_vl_text",
|
| 25 |
+
"num_attention_heads": 32,
|
| 26 |
+
"num_hidden_layers": 36,
|
| 27 |
+
"num_key_value_heads": 8,
|
| 28 |
+
"pad_token_id": null,
|
| 29 |
+
"rms_norm_eps": 1e-06,
|
| 30 |
+
"rope_parameters": {
|
| 31 |
+
"mrope_interleaved": true,
|
| 32 |
+
"mrope_section": [
|
| 33 |
+
24,
|
| 34 |
+
20,
|
| 35 |
+
20
|
| 36 |
+
],
|
| 37 |
+
"rope_theta": 5000000,
|
| 38 |
+
"rope_type": "default"
|
| 39 |
+
},
|
| 40 |
+
"use_cache": false,
|
| 41 |
+
"vocab_size": 151936
|
| 42 |
+
},
|
| 43 |
+
"tie_word_embeddings": false,
|
| 44 |
+
"transformers_version": "5.8.0",
|
| 45 |
+
"use_cache": false,
|
| 46 |
+
"video_token_id": 151656,
|
| 47 |
+
"vision_config": {
|
| 48 |
+
"deepstack_visual_indexes": [
|
| 49 |
+
8,
|
| 50 |
+
16,
|
| 51 |
+
24
|
| 52 |
+
],
|
| 53 |
+
"depth": 27,
|
| 54 |
+
"dtype": "bfloat16",
|
| 55 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 56 |
+
"hidden_size": 1152,
|
| 57 |
+
"in_channels": 3,
|
| 58 |
+
"initializer_range": 0.02,
|
| 59 |
+
"intermediate_size": 4304,
|
| 60 |
+
"model_type": "qwen3_vl_vision",
|
| 61 |
+
"num_heads": 16,
|
| 62 |
+
"num_position_embeddings": 2304,
|
| 63 |
+
"out_hidden_size": 4096,
|
| 64 |
+
"patch_size": 16,
|
| 65 |
+
"spatial_merge_size": 2,
|
| 66 |
+
"temporal_patch_size": 2
|
| 67 |
+
},
|
| 68 |
+
"vision_end_token_id": 151653,
|
| 69 |
+
"vision_start_token_id": 151652
|
| 70 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/eval_results.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.1253263707571803,
|
| 3 |
+
"eval_loss": 0.08317559212446213,
|
| 4 |
+
"eval_runtime": 23.7677,
|
| 5 |
+
"eval_samples_per_second": 13.59,
|
| 6 |
+
"eval_steps_per_second": 3.408
|
| 7 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_sample": true,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
151645,
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"repetition_penalty": 1.0,
|
| 10 |
+
"temperature": 0.7,
|
| 11 |
+
"top_k": 20,
|
| 12 |
+
"top_p": 0.8,
|
| 13 |
+
"transformers_version": "5.8.0"
|
| 14 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd98a2902f0e9254afeb035fc7adc3f7aa7a37a9c803745729d8fbbe53a1d7d1
|
| 3 |
+
size 17534340584
|
grail_v1_343_b64_sft_e3_ck300/processor_config.json
ADDED
|
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_processor": {
|
| 3 |
+
"do_convert_rgb": true,
|
| 4 |
+
"do_normalize": true,
|
| 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": "Qwen2VLImageProcessor",
|
| 13 |
+
"image_std": [
|
| 14 |
+
0.5,
|
| 15 |
+
0.5,
|
| 16 |
+
0.5
|
| 17 |
+
],
|
| 18 |
+
"merge_size": 2,
|
| 19 |
+
"patch_size": 16,
|
| 20 |
+
"resample": 3,
|
| 21 |
+
"rescale_factor": 0.00392156862745098,
|
| 22 |
+
"size": {
|
| 23 |
+
"longest_edge": 16777216,
|
| 24 |
+
"shortest_edge": 65536
|
| 25 |
+
},
|
| 26 |
+
"temporal_patch_size": 2
|
| 27 |
+
},
|
| 28 |
+
"processor_class": "Qwen3VLProcessor",
|
| 29 |
+
"video_processor": {
|
| 30 |
+
"do_convert_rgb": true,
|
| 31 |
+
"do_normalize": true,
|
| 32 |
+
"do_rescale": true,
|
| 33 |
+
"do_resize": true,
|
| 34 |
+
"do_sample_frames": true,
|
| 35 |
+
"fps": 2,
|
| 36 |
+
"image_mean": [
|
| 37 |
+
0.5,
|
| 38 |
+
0.5,
|
| 39 |
+
0.5
|
| 40 |
+
],
|
| 41 |
+
"image_std": [
|
| 42 |
+
0.5,
|
| 43 |
+
0.5,
|
| 44 |
+
0.5
|
| 45 |
+
],
|
| 46 |
+
"max_frames": 768,
|
| 47 |
+
"merge_size": 2,
|
| 48 |
+
"min_frames": 4,
|
| 49 |
+
"patch_size": 16,
|
| 50 |
+
"resample": 3,
|
| 51 |
+
"rescale_factor": 0.00392156862745098,
|
| 52 |
+
"return_metadata": false,
|
| 53 |
+
"size": {
|
| 54 |
+
"longest_edge": 25165824,
|
| 55 |
+
"shortest_edge": 4096
|
| 56 |
+
},
|
| 57 |
+
"temporal_patch_size": 2,
|
| 58 |
+
"video_processor_type": "Qwen3VLVideoProcessor"
|
| 59 |
+
}
|
| 60 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/sft.yaml
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# LLaMA-Factory full-parameter SFT of Qwen3-VL-8B-Instruct on one HER-final dataset.
|
| 2 |
+
# Four L40S (46 GB), one run at a time: DeepSpeed ZeRO-3 with optimizer offload to host
|
| 3 |
+
# RAM (~130 GB per run, so two runs do not fit in 372 GB); the vision tower and projector
|
| 4 |
+
# stay frozen (LLaMA-Factory's default for VLMs), the LLM trains.
|
| 5 |
+
# Filled per run by launch_sft.sh: grail_v1_343_sft_e3, exps/sft/grail_v1_343_sft_e3_k1536, /scratch/kwc5354/viewagent/viewagent_private/exps/sft_runs/grail_v1_343_b64_sft_e3.
|
| 6 |
+
### model
|
| 7 |
+
model_name_or_path: /scratch/kwc5354/viewagent/viewagent_private/exps/models/graph_nemi_sft_e2
|
| 8 |
+
template: qwen3_vl_nothink
|
| 9 |
+
trust_remote_code: true
|
| 10 |
+
image_max_pixels: 65536 # frames are 256x256 already; never upscale
|
| 11 |
+
image_min_pixels: 4096
|
| 12 |
+
|
| 13 |
+
### method
|
| 14 |
+
stage: sft
|
| 15 |
+
do_train: true
|
| 16 |
+
finetuning_type: full
|
| 17 |
+
freeze_vision_tower: true
|
| 18 |
+
freeze_multi_modal_projector: true
|
| 19 |
+
freeze_language_model: false
|
| 20 |
+
deepspeed: /scratch/kwc5354/viewagent/viewagent_private/LLaMA-Factory/examples/deepspeed/ds_z3_config.json
|
| 21 |
+
|
| 22 |
+
### dataset
|
| 23 |
+
dataset: grail_v1_343_sft_e3
|
| 24 |
+
dataset_dir: exps/sft/grail_v1_343_sft_e3_k1536
|
| 25 |
+
media_dir: /scratch/kwc5354/viewagent/viewagent_private
|
| 26 |
+
cutoff_len: 8192
|
| 27 |
+
overwrite_cache: true
|
| 28 |
+
preprocessing_num_workers: 16
|
| 29 |
+
dataloader_num_workers: 4
|
| 30 |
+
|
| 31 |
+
### output
|
| 32 |
+
output_dir: /scratch/kwc5354/viewagent/viewagent_private/exps/sft_runs/grail_v1_343_b64_sft_e3
|
| 33 |
+
logging_steps: 10
|
| 34 |
+
save_strategy: steps
|
| 35 |
+
save_steps: 100
|
| 36 |
+
save_total_limit: 100
|
| 37 |
+
save_only_model: true
|
| 38 |
+
plot_loss: true
|
| 39 |
+
overwrite_output_dir: true
|
| 40 |
+
report_to: wandb
|
| 41 |
+
|
| 42 |
+
### train
|
| 43 |
+
per_device_train_batch_size: 1
|
| 44 |
+
gradient_accumulation_steps: 16
|
| 45 |
+
max_steps: 300
|
| 46 |
+
learning_rate: 1.0e-5
|
| 47 |
+
lr_scheduler_type: cosine
|
| 48 |
+
warmup_ratio: 0.03
|
| 49 |
+
weight_decay: 0.01
|
| 50 |
+
bf16: true
|
| 51 |
+
gradient_checkpointing: true
|
| 52 |
+
ddp_timeout: 180000000
|
| 53 |
+
seed: 0
|
| 54 |
+
val_size: 0.05
|
| 55 |
+
eval_strategy: steps
|
| 56 |
+
eval_steps: 100
|
| 57 |
+
per_device_eval_batch_size: 1
|
| 58 |
+
run_name: grail_v1_343_b64_sft_e3
|
grail_v1_343_b64_sft_e3_ck300/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
|
| 3 |
+
size 11422650
|
grail_v1_343_b64_sft_e3_ck300/tokenizer_config.json
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"bos_token": null,
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eos_token": "<|im_end|>",
|
| 7 |
+
"errors": "replace",
|
| 8 |
+
"extra_special_tokens": [
|
| 9 |
+
"<|im_start|>",
|
| 10 |
+
"<|im_end|>",
|
| 11 |
+
"<|object_ref_start|>",
|
| 12 |
+
"<|object_ref_end|>",
|
| 13 |
+
"<|box_start|>",
|
| 14 |
+
"<|box_end|>",
|
| 15 |
+
"<|quad_start|>",
|
| 16 |
+
"<|quad_end|>",
|
| 17 |
+
"<|vision_start|>",
|
| 18 |
+
"<|vision_end|>",
|
| 19 |
+
"<|vision_pad|>",
|
| 20 |
+
"<|image_pad|>",
|
| 21 |
+
"<|video_pad|>"
|
| 22 |
+
],
|
| 23 |
+
"is_local": true,
|
| 24 |
+
"local_files_only": false,
|
| 25 |
+
"model_max_length": 262144,
|
| 26 |
+
"pad_token": "<|endoftext|>",
|
| 27 |
+
"padding_side": "right",
|
| 28 |
+
"processor_class": "Qwen3VLProcessor",
|
| 29 |
+
"split_special_tokens": false,
|
| 30 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 31 |
+
"unk_token": null
|
| 32 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.1253263707571803,
|
| 3 |
+
"total_flos": 138960941613056.0,
|
| 4 |
+
"train_loss": 0.12055180182059606,
|
| 5 |
+
"train_runtime": 3553.0676,
|
| 6 |
+
"train_samples_per_second": 5.404,
|
| 7 |
+
"train_steps_per_second": 0.084
|
| 8 |
+
}
|
grail_v1_343_b64_sft_e3_ck300/trainer_state.json
ADDED
|
@@ -0,0 +1,277 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 3.1253263707571803,
|
| 6 |
+
"eval_steps": 100,
|
| 7 |
+
"global_step": 300,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.10443864229765012,
|
| 14 |
+
"grad_norm": 3.4800231617558675,
|
| 15 |
+
"learning_rate": 1e-05,
|
| 16 |
+
"loss": 0.9193525314331055,
|
| 17 |
+
"step": 10
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.20887728459530025,
|
| 21 |
+
"grad_norm": 0.7899235139957301,
|
| 22 |
+
"learning_rate": 9.970890689061622e-06,
|
| 23 |
+
"loss": 0.2244274377822876,
|
| 24 |
+
"step": 20
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.3133159268929504,
|
| 28 |
+
"grad_norm": 0.617801418499001,
|
| 29 |
+
"learning_rate": 9.883901697039809e-06,
|
| 30 |
+
"loss": 0.16480841636657714,
|
| 31 |
+
"step": 30
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 0.4177545691906005,
|
| 35 |
+
"grad_norm": 0.48127310176681687,
|
| 36 |
+
"learning_rate": 9.740045899781353e-06,
|
| 37 |
+
"loss": 0.15318325757980347,
|
| 38 |
+
"step": 40
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.5221932114882507,
|
| 42 |
+
"grad_norm": 0.49629249618330495,
|
| 43 |
+
"learning_rate": 9.540998314539327e-06,
|
| 44 |
+
"loss": 0.13782763481140137,
|
| 45 |
+
"step": 50
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.6266318537859008,
|
| 49 |
+
"grad_norm": 0.4770197103054743,
|
| 50 |
+
"learning_rate": 9.289076596533873e-06,
|
| 51 |
+
"loss": 0.13069230318069458,
|
| 52 |
+
"step": 60
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"epoch": 0.7310704960835509,
|
| 56 |
+
"grad_norm": 0.47581632537537943,
|
| 57 |
+
"learning_rate": 8.987214052813605e-06,
|
| 58 |
+
"loss": 0.12325828075408936,
|
| 59 |
+
"step": 70
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 0.835509138381201,
|
| 63 |
+
"grad_norm": 0.4927505861945296,
|
| 64 |
+
"learning_rate": 8.638925487636847e-06,
|
| 65 |
+
"loss": 0.12904804944992065,
|
| 66 |
+
"step": 80
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"epoch": 0.9399477806788512,
|
| 70 |
+
"grad_norm": 0.521338609802959,
|
| 71 |
+
"learning_rate": 8.248266277059607e-06,
|
| 72 |
+
"loss": 0.12003054618835449,
|
| 73 |
+
"step": 90
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 1.04177545691906,
|
| 77 |
+
"grad_norm": 0.6308363302110879,
|
| 78 |
+
"learning_rate": 7.819785149254534e-06,
|
| 79 |
+
"loss": 0.1107589602470398,
|
| 80 |
+
"step": 100
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 1.04177545691906,
|
| 84 |
+
"eval_loss": 0.1119612380862236,
|
| 85 |
+
"eval_runtime": 24.1511,
|
| 86 |
+
"eval_samples_per_second": 13.374,
|
| 87 |
+
"eval_steps_per_second": 3.354,
|
| 88 |
+
"step": 100
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1462140992167102,
|
| 92 |
+
"grad_norm": 0.5093325690718998,
|
| 93 |
+
"learning_rate": 7.358471220373831e-06,
|
| 94 |
+
"loss": 0.1009824275970459,
|
| 95 |
+
"step": 110
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2506527415143602,
|
| 99 |
+
"grad_norm": 0.5165850281407918,
|
| 100 |
+
"learning_rate": 6.869695902655898e-06,
|
| 101 |
+
"loss": 0.10302877426147461,
|
| 102 |
+
"step": 120
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3550913838120104,
|
| 106 |
+
"grad_norm": 0.6941090040542808,
|
| 107 |
+
"learning_rate": 6.3591503611817155e-06,
|
| 108 |
+
"loss": 0.09875673651695252,
|
| 109 |
+
"step": 130
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4595300261096606,
|
| 113 |
+
"grad_norm": 0.5615339236677334,
|
| 114 |
+
"learning_rate": 5.832779247517273e-06,
|
| 115 |
+
"loss": 0.09152050614356995,
|
| 116 |
+
"step": 140
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5639686684073109,
|
| 120 |
+
"grad_norm": 0.5648506068607166,
|
| 121 |
+
"learning_rate": 5.296711481829227e-06,
|
| 122 |
+
"loss": 0.09409712553024292,
|
| 123 |
+
"step": 150
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6684073107049608,
|
| 127 |
+
"grad_norm": 0.5782574926659386,
|
| 128 |
+
"learning_rate": 4.757188889427761e-06,
|
| 129 |
+
"loss": 0.08830703496932983,
|
| 130 |
+
"step": 160
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.7728459530026108,
|
| 134 |
+
"grad_norm": 0.5395237444164678,
|
| 135 |
+
"learning_rate": 4.220493522673067e-06,
|
| 136 |
+
"loss": 0.08994704484939575,
|
| 137 |
+
"step": 170
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.877284595300261,
|
| 141 |
+
"grad_norm": 0.560459804746689,
|
| 142 |
+
"learning_rate": 3.6928745144891733e-06,
|
| 143 |
+
"loss": 0.08162439465522767,
|
| 144 |
+
"step": 180
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.9817232375979112,
|
| 148 |
+
"grad_norm": 0.8105973003467271,
|
| 149 |
+
"learning_rate": 3.180475315182563e-06,
|
| 150 |
+
"loss": 0.07880358695983887,
|
| 151 |
+
"step": 190
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.08355091383812,
|
| 155 |
+
"grad_norm": 0.5902861780112129,
|
| 156 |
+
"learning_rate": 2.6892621598001157e-06,
|
| 157 |
+
"loss": 0.06284728050231933,
|
| 158 |
+
"step": 200
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 2.08355091383812,
|
| 162 |
+
"eval_loss": 0.087026447057724,
|
| 163 |
+
"eval_runtime": 23.9662,
|
| 164 |
+
"eval_samples_per_second": 13.477,
|
| 165 |
+
"eval_steps_per_second": 3.38,
|
| 166 |
+
"step": 200
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"epoch": 2.1879895561357703,
|
| 170 |
+
"grad_norm": 0.7554679320867941,
|
| 171 |
+
"learning_rate": 2.2249545989326516e-06,
|
| 172 |
+
"loss": 0.054855984449386594,
|
| 173 |
+
"step": 210
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"epoch": 2.2924281984334205,
|
| 177 |
+
"grad_norm": 0.7398042667879547,
|
| 178 |
+
"learning_rate": 1.7929589018443016e-06,
|
| 179 |
+
"loss": 0.05535943508148193,
|
| 180 |
+
"step": 220
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"epoch": 2.3968668407310707,
|
| 184 |
+
"grad_norm": 0.6432109525498831,
|
| 185 |
+
"learning_rate": 1.3983051073632996e-06,
|
| 186 |
+
"loss": 0.0549279510974884,
|
| 187 |
+
"step": 230
|
| 188 |
+
},
|
| 189 |
+
{
|
| 190 |
+
"epoch": 2.5013054830287205,
|
| 191 |
+
"grad_norm": 0.7164340031047494,
|
| 192 |
+
"learning_rate": 1.0455884554962725e-06,
|
| 193 |
+
"loss": 0.05147741436958313,
|
| 194 |
+
"step": 240
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 2.6057441253263707,
|
| 198 |
+
"grad_norm": 0.890018499051114,
|
| 199 |
+
"learning_rate": 7.389158817201541e-07,
|
| 200 |
+
"loss": 0.050784957408905027,
|
| 201 |
+
"step": 250
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 2.710182767624021,
|
| 205 |
+
"grad_norm": 0.7126411786716337,
|
| 206 |
+
"learning_rate": 4.818581969574743e-07,
|
| 207 |
+
"loss": 0.05355076789855957,
|
| 208 |
+
"step": 260
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"epoch": 2.814621409921671,
|
| 212 |
+
"grad_norm": 0.7363642785039943,
|
| 213 |
+
"learning_rate": 2.774085100381735e-07,
|
| 214 |
+
"loss": 0.05100441575050354,
|
| 215 |
+
"step": 270
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 2.9190600522193213,
|
| 219 |
+
"grad_norm": 0.7662362067099162,
|
| 220 |
+
"learning_rate": 1.2794737676536993e-07,
|
| 221 |
+
"loss": 0.05301822423934936,
|
| 222 |
+
"step": 280
|
| 223 |
+
},
|
| 224 |
+
{
|
| 225 |
+
"epoch": 3.02088772845953,
|
| 226 |
+
"grad_norm": 0.47799321860734684,
|
| 227 |
+
"learning_rate": 3.521508137971807e-08,
|
| 228 |
+
"loss": 0.04839288592338562,
|
| 229 |
+
"step": 290
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"epoch": 3.1253263707571803,
|
| 233 |
+
"grad_norm": 0.4948629904757914,
|
| 234 |
+
"learning_rate": 2.9137316938265826e-10,
|
| 235 |
+
"loss": 0.03987968862056732,
|
| 236 |
+
"step": 300
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 3.1253263707571803,
|
| 240 |
+
"eval_loss": 0.08317559212446213,
|
| 241 |
+
"eval_runtime": 23.9065,
|
| 242 |
+
"eval_samples_per_second": 13.511,
|
| 243 |
+
"eval_steps_per_second": 3.388,
|
| 244 |
+
"step": 300
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"epoch": 3.1253263707571803,
|
| 248 |
+
"step": 300,
|
| 249 |
+
"total_flos": 138960941613056.0,
|
| 250 |
+
"train_loss": 0.12055180182059606,
|
| 251 |
+
"train_runtime": 3553.0676,
|
| 252 |
+
"train_samples_per_second": 5.404,
|
| 253 |
+
"train_steps_per_second": 0.084
|
| 254 |
+
}
|
| 255 |
+
],
|
| 256 |
+
"logging_steps": 10,
|
| 257 |
+
"max_steps": 300,
|
| 258 |
+
"num_input_tokens_seen": 0,
|
| 259 |
+
"num_train_epochs": 4,
|
| 260 |
+
"save_steps": 100,
|
| 261 |
+
"stateful_callbacks": {
|
| 262 |
+
"TrainerControl": {
|
| 263 |
+
"args": {
|
| 264 |
+
"should_epoch_stop": false,
|
| 265 |
+
"should_evaluate": false,
|
| 266 |
+
"should_log": false,
|
| 267 |
+
"should_save": true,
|
| 268 |
+
"should_training_stop": true
|
| 269 |
+
},
|
| 270 |
+
"attributes": {}
|
| 271 |
+
}
|
| 272 |
+
},
|
| 273 |
+
"total_flos": 138960941613056.0,
|
| 274 |
+
"train_batch_size": 1,
|
| 275 |
+
"trial_name": null,
|
| 276 |
+
"trial_params": null
|
| 277 |
+
}
|