From b9263c656eb6b901bf797dc105eb5c12d1a2442d Mon Sep 17 00:00:00 2001 From: paramthakkar123 Date: Tue, 31 Mar 2026 00:24:06 +0530 Subject: [PATCH 1/5] Added Distilkit task --- tasks/8_distilkit-distillation.json | 49 +++++++++++++++++++++++++++++ 1 file changed, 49 insertions(+) create mode 100644 tasks/8_distilkit-distillation.json diff --git a/tasks/8_distilkit-distillation.json b/tasks/8_distilkit-distillation.json new file mode 100644 index 0000000..8e820b0 --- /dev/null +++ b/tasks/8_distilkit-distillation.json @@ -0,0 +1,49 @@ +[ + { + "title": "DistilKit Distillation", + "description": "A task for distilling large language models using DistilKit. Trains a student model using knowledge distillation from a teacher model or dataset.", + "github_repo_url": "https://github.com/transformerlab/transformerlab-examples", + "github_repo_dir": "distilkit-distillation", + "setup": "git clone https://github.com/arcee-ai/distillkit.git && cd distillkit && pip install -e . && uv pip install transformers datasets huggingface-hub torch wandb", + "run": "python ~/distilkit-distillation/train.py", + "envs": { + "WANDB_PROJECT": "distilkit-distillation", + "PYTHONUNBUFFERED": "1", + "HF_TOKEN": "{{secret._HF_TOKEN}}", + "WANDB_API_KEY": "{{secret._WANDB_API_KEY}}" + }, + "parameters": { + "model": "Qwen/Qwen3-8B", + "output_path": "./output", + "sequence_length": 8192, + "train_dataset_repo": "arcee-ai/Qwen3-235B-Logits-Packed-8192", + "train_dataset_split": "train", + "teacher_kind": "dataset", + "vocab_size": 151936, + "k": 128, + "exact_k": 32, + "num_train_epochs": 1, + "per_device_train_batch_size": 1, + "gradient_accumulation_steps": 8, + "learning_rate": 0.000002, + "bf16": true, + "optim": "adamw_torch", + "gradient_checkpointing": true + }, + "metadata": { + "category": "training", + "modality": "text", + "framework": [ + "distilkit", + "huggingface" + ] + }, + "supportedAccelerators": { + "NVIDIA": { + "resources": { + "accelerators": "A100:1" + } + } + } + } +] From ba9fe047f8d6c9a8397ac249a1fd3f49bcd39daa Mon Sep 17 00:00:00 2001 From: paramthakkar123 Date: Tue, 31 Mar 2026 00:28:30 +0530 Subject: [PATCH 2/5] Modified video diffusion task and distilkit params --- tasks/7_video_diffusion.json | 19 ++++++++++++++++++- tasks/8_distilkit-distillation.json | 1 - 2 files changed, 18 insertions(+), 2 deletions(-) diff --git a/tasks/7_video_diffusion.json b/tasks/7_video_diffusion.json index 807d028..19c1a33 100644 --- a/tasks/7_video_diffusion.json +++ b/tasks/7_video_diffusion.json @@ -1,5 +1,6 @@ [ { + "name": "video-generation-wan-task", "title": "Video Generation with Wan2.1", "description": "A video generation task using the Wan2.1-T2V model. Generates videos from text prompts using the Wan pipeline from Diffusers. Requires HF_TOKEN environment variable for model access.", "github_repo_url": "https://github.com/transformerlab/transformerlab-examples", @@ -17,6 +18,22 @@ "accelerators": "A100:1" } } - } + }, + "envs": { + "PYTHONUNBUFFERED": "1", + "HF_TOKEN": "{{secret._HF_TOKEN}}" + }, + "setup": "uv pip install --upgrade pip && uv pip install git+https://github.com/huggingface/diffusers.git && uv pip install 'transformers>=4.41.0' 'peft>=0.17.0' accelerate ftfy huggingface-hub opencv-python transformerlab imageio-ffmpeg", + "parameters": { + "model_id": "Wan-AI/Wan2.1-T2V-1.3B-Diffusers", + "prompt": "A cat walks on the grass, realistic", + "negative_prompt": "", + "height": 480, + "width": 832, + "num_frames": 81, + "guidance_scale": 5.0, + "fps": 15 + }, + "run": "python ~/video-diffusion/main.py" } ] diff --git a/tasks/8_distilkit-distillation.json b/tasks/8_distilkit-distillation.json index 8e820b0..6c10bba 100644 --- a/tasks/8_distilkit-distillation.json +++ b/tasks/8_distilkit-distillation.json @@ -14,7 +14,6 @@ }, "parameters": { "model": "Qwen/Qwen3-8B", - "output_path": "./output", "sequence_length": 8192, "train_dataset_repo": "arcee-ai/Qwen3-235B-Logits-Packed-8192", "train_dataset_split": "train", From 4fd46328d1b4023baf40a18b4b45890637b7c463 Mon Sep 17 00:00:00 2001 From: paramthakkar123 Date: Tue, 31 Mar 2026 00:35:03 +0530 Subject: [PATCH 3/5] Removed params and envs from gallery --- tasks/7_video_diffusion.json | 14 -------------- tasks/8_distilkit-distillation.json | 23 ----------------------- 2 files changed, 37 deletions(-) diff --git a/tasks/7_video_diffusion.json b/tasks/7_video_diffusion.json index 19c1a33..5cee9de 100644 --- a/tasks/7_video_diffusion.json +++ b/tasks/7_video_diffusion.json @@ -19,21 +19,7 @@ } } }, - "envs": { - "PYTHONUNBUFFERED": "1", - "HF_TOKEN": "{{secret._HF_TOKEN}}" - }, "setup": "uv pip install --upgrade pip && uv pip install git+https://github.com/huggingface/diffusers.git && uv pip install 'transformers>=4.41.0' 'peft>=0.17.0' accelerate ftfy huggingface-hub opencv-python transformerlab imageio-ffmpeg", - "parameters": { - "model_id": "Wan-AI/Wan2.1-T2V-1.3B-Diffusers", - "prompt": "A cat walks on the grass, realistic", - "negative_prompt": "", - "height": 480, - "width": 832, - "num_frames": 81, - "guidance_scale": 5.0, - "fps": 15 - }, "run": "python ~/video-diffusion/main.py" } ] diff --git a/tasks/8_distilkit-distillation.json b/tasks/8_distilkit-distillation.json index 6c10bba..50f8325 100644 --- a/tasks/8_distilkit-distillation.json +++ b/tasks/8_distilkit-distillation.json @@ -6,29 +6,6 @@ "github_repo_dir": "distilkit-distillation", "setup": "git clone https://github.com/arcee-ai/distillkit.git && cd distillkit && pip install -e . && uv pip install transformers datasets huggingface-hub torch wandb", "run": "python ~/distilkit-distillation/train.py", - "envs": { - "WANDB_PROJECT": "distilkit-distillation", - "PYTHONUNBUFFERED": "1", - "HF_TOKEN": "{{secret._HF_TOKEN}}", - "WANDB_API_KEY": "{{secret._WANDB_API_KEY}}" - }, - "parameters": { - "model": "Qwen/Qwen3-8B", - "sequence_length": 8192, - "train_dataset_repo": "arcee-ai/Qwen3-235B-Logits-Packed-8192", - "train_dataset_split": "train", - "teacher_kind": "dataset", - "vocab_size": 151936, - "k": 128, - "exact_k": 32, - "num_train_epochs": 1, - "per_device_train_batch_size": 1, - "gradient_accumulation_steps": 8, - "learning_rate": 0.000002, - "bf16": true, - "optim": "adamw_torch", - "gradient_checkpointing": true - }, "metadata": { "category": "training", "modality": "text", From 717a622dfb039b7fdcfa905c0c3bcdab8363f7cf Mon Sep 17 00:00:00 2001 From: paramthakkar123 Date: Tue, 31 Mar 2026 00:36:06 +0530 Subject: [PATCH 4/5] Removed params and envs from gallery --- tasks/7_video_diffusion.json | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/tasks/7_video_diffusion.json b/tasks/7_video_diffusion.json index 5cee9de..39d7a7c 100644 --- a/tasks/7_video_diffusion.json +++ b/tasks/7_video_diffusion.json @@ -1,6 +1,5 @@ [ { - "name": "video-generation-wan-task", "title": "Video Generation with Wan2.1", "description": "A video generation task using the Wan2.1-T2V model. Generates videos from text prompts using the Wan pipeline from Diffusers. Requires HF_TOKEN environment variable for model access.", "github_repo_url": "https://github.com/transformerlab/transformerlab-examples", @@ -18,8 +17,6 @@ "accelerators": "A100:1" } } - }, - "setup": "uv pip install --upgrade pip && uv pip install git+https://github.com/huggingface/diffusers.git && uv pip install 'transformers>=4.41.0' 'peft>=0.17.0' accelerate ftfy huggingface-hub opencv-python transformerlab imageio-ffmpeg", - "run": "python ~/video-diffusion/main.py" + } } -] +] \ No newline at end of file From fc75a3716680ac2792716c65607becf3110b6782 Mon Sep 17 00:00:00 2001 From: paramthakkar123 Date: Tue, 31 Mar 2026 00:36:37 +0530 Subject: [PATCH 5/5] Updates --- tasks/7_video_diffusion.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tasks/7_video_diffusion.json b/tasks/7_video_diffusion.json index 39d7a7c..807d028 100644 --- a/tasks/7_video_diffusion.json +++ b/tasks/7_video_diffusion.json @@ -19,4 +19,4 @@ } } } -] \ No newline at end of file +]