From b88c71e00c24be590f55c1b71cd054e081689e1e Mon Sep 17 00:00:00 2001 From: Jake Poznanski Date: Wed, 16 Jul 2025 20:41:35 +0000 Subject: [PATCH] Rounding to better image size, full soups --- ..._soupfull0.yaml => qwen25_vl_olmocrv2_1288_soupfull0.yaml} | 4 ++-- ..._soupfull1.yaml => qwen25_vl_olmocrv2_1288_soupfull1.yaml} | 4 ++-- ..._soupfull2.yaml => qwen25_vl_olmocrv2_1288_soupfull2.yaml} | 4 ++-- 3 files changed, 6 insertions(+), 6 deletions(-) rename olmocr/train/configs/{qwen25_vl_olmocrv2_soupfull0.yaml => qwen25_vl_olmocrv2_1288_soupfull0.yaml} (96%) rename olmocr/train/configs/{qwen25_vl_olmocrv2_soupfull1.yaml => qwen25_vl_olmocrv2_1288_soupfull1.yaml} (96%) rename olmocr/train/configs/{qwen25_vl_olmocrv2_soupfull2.yaml => qwen25_vl_olmocrv2_1288_soupfull2.yaml} (96%) diff --git a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull0.yaml b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull0.yaml similarity index 96% rename from olmocr/train/configs/qwen25_vl_olmocrv2_soupfull0.yaml rename to olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull0.yaml index 42387b7..be76a72 100644 --- a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull0.yaml +++ b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull0.yaml @@ -2,7 +2,7 @@ # Project metadata project_name: olmocr-qwen-vl-training -run_name: qwen2.5-vl-7b-olmocrv2_soupfull0 +run_name: qwen2.5-vl-7b-olmocrv2-1288-soupfull0 # Model configuration model: @@ -33,7 +33,7 @@ dataset: - name: FrontMatterParser front_matter_class: PageResponse - name: PDFRenderer - target_longest_image_dim: 1280 + target_longest_image_dim: 1288 - name: NewYamlFinetuningPromptWithNoAnchoring - name: FrontMatterOutputFormat - name: InstructUserMessages diff --git a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull1.yaml b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull1.yaml similarity index 96% rename from olmocr/train/configs/qwen25_vl_olmocrv2_soupfull1.yaml rename to olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull1.yaml index dd40986..11192d6 100644 --- a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull1.yaml +++ b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull1.yaml @@ -2,7 +2,7 @@ # Project metadata project_name: olmocr-qwen-vl-training -run_name: qwen2.5-vl-7b-olmocrv2_soupfull1 +run_name: qwen2.5-vl-7b-olmocrv2-1288-soupfull1 # Model configuration model: @@ -33,7 +33,7 @@ dataset: - name: FrontMatterParser front_matter_class: PageResponse - name: PDFRenderer - target_longest_image_dim: 1280 + target_longest_image_dim: 1288 - name: NewYamlFinetuningPromptWithNoAnchoring - name: FrontMatterOutputFormat - name: InstructUserMessages diff --git a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull2.yaml b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull2.yaml similarity index 96% rename from olmocr/train/configs/qwen25_vl_olmocrv2_soupfull2.yaml rename to olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull2.yaml index 0214cee..3d04d44 100644 --- a/olmocr/train/configs/qwen25_vl_olmocrv2_soupfull2.yaml +++ b/olmocr/train/configs/qwen25_vl_olmocrv2_1288_soupfull2.yaml @@ -2,7 +2,7 @@ # Project metadata project_name: olmocr-qwen-vl-training -run_name: qwen2.5-vl-7b-olmocrv2_soupfull1 +run_name: qwen2.5-vl-7b-olmocrv2-1288-soupfull2 # Model configuration model: @@ -33,7 +33,7 @@ dataset: - name: FrontMatterParser front_matter_class: PageResponse - name: PDFRenderer - target_longest_image_dim: 1280 + target_longest_image_dim: 1288 - name: NewYamlFinetuningPromptWithNoAnchoring - name: FrontMatterOutputFormat - name: InstructUserMessages