From f47dbb9a1ea778332b40f867885f6559fd656600 Mon Sep 17 00:00:00 2001 From: degenfabian Date: Tue, 19 Aug 2025 02:50:46 +0200 Subject: [PATCH 1/4] updating loading in stable lm demo to use transformer bridge --- demos/stable_lm.ipynb | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/demos/stable_lm.ipynb b/demos/stable_lm.ipynb index bfe623c36..0f70be25f 100644 --- a/demos/stable_lm.ipynb +++ b/demos/stable_lm.ipynb @@ -60,21 +60,21 @@ }, { "cell_type": "code", - "execution_count": 2, + "execution_count": null, "metadata": { "id": "PXB6xkimoH2h" }, "outputs": [], "source": [ "import torch\n", - "from transformer_lens import HookedTransformer\n", + "from transformer_lens.model_bridge import TransformerBridge\n", "\n", "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"" ] }, { "cell_type": "code", - "execution_count": 3, + "execution_count": null, "metadata": { "colab": { "base_uri": "https://localhost:8080/", @@ -355,7 +355,8 @@ "source": [ "# Load the 3 billion parameters version in 16 bits\n", "# You can increase the precision or the size if you have enough GPU RAM available\n", - "model = HookedTransformer.from_pretrained(\"stabilityai/stablelm-tuned-alpha-3b\", torch_dtype=torch.bfloat16, device=device)" + "model = TransformerBridge.boot_transformers(\"stabilityai/stablelm-tuned-alpha-3b\", torch_dtype=torch.bfloat16, device=device)\n", + "model.enable_compatibility_mode()" ] }, { From 8a7f3a6fccc5c9033a4d25dc5a4e7da0fd7a3608 Mon Sep 17 00:00:00 2001 From: degenfabian Date: Tue, 19 Aug 2025 18:01:25 +0200 Subject: [PATCH 2/4] add stable_lm demo to CI --- .github/workflows/checks.yml | 1 + 1 file changed, 1 insertion(+) diff --git a/.github/workflows/checks.yml b/.github/workflows/checks.yml index 4de51026c..d3e147268 100644 --- a/.github/workflows/checks.yml +++ b/.github/workflows/checks.yml @@ -159,6 +159,7 @@ jobs: # - "No_Position_Experiment" - "Othello_GPT" - "Patchscopes_Generation_Demo" + - "stable_lm" # - "T5" steps: - uses: actions/checkout@v3 From 2be7624add18de8fd48403e372e3f2e28fe7aafa Mon Sep 17 00:00:00 2001 From: Bryce Meyer Date: Thu, 16 Oct 2025 12:17:50 +0200 Subject: [PATCH 3/4] fixed booting --- demos/stable_lm.ipynb | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/demos/stable_lm.ipynb b/demos/stable_lm.ipynb index 0f70be25f..75a6523d9 100644 --- a/demos/stable_lm.ipynb +++ b/demos/stable_lm.ipynb @@ -355,7 +355,7 @@ "source": [ "# Load the 3 billion parameters version in 16 bits\n", "# You can increase the precision or the size if you have enough GPU RAM available\n", - "model = TransformerBridge.boot_transformers(\"stabilityai/stablelm-tuned-alpha-3b\", torch_dtype=torch.bfloat16, device=device)\n", + "model = TransformerBridge.boot_transformers(\"stabilityai/stablelm-tuned-alpha-3b\", dtype=torch.bfloat16, device=device)\n", "model.enable_compatibility_mode()" ] }, From 15a1e48bcc9f2e85345177ef11926a54c2d95c9c Mon Sep 17 00:00:00 2001 From: Bryce Meyer Date: Thu, 16 Oct 2025 20:31:55 +0200 Subject: [PATCH 4/4] added dtype selction --- demos/stable_lm.ipynb | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/demos/stable_lm.ipynb b/demos/stable_lm.ipynb index 75a6523d9..f8fd37b00 100644 --- a/demos/stable_lm.ipynb +++ b/demos/stable_lm.ipynb @@ -69,7 +69,8 @@ "import torch\n", "from transformer_lens.model_bridge import TransformerBridge\n", "\n", - "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"" + "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n", + "dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32" ] }, { @@ -355,7 +356,7 @@ "source": [ "# Load the 3 billion parameters version in 16 bits\n", "# You can increase the precision or the size if you have enough GPU RAM available\n", - "model = TransformerBridge.boot_transformers(\"stabilityai/stablelm-tuned-alpha-3b\", dtype=torch.bfloat16, device=device)\n", + "model = TransformerBridge.boot_transformers(\"stabilityai/stablelm-tuned-alpha-3b\", dtype=dtype, device=device)\n", "model.enable_compatibility_mode()" ] },