[object Object]

← back to Exo

add deepseek-v3-3bit and deepseek-r1-3bit

82ef08601049be3f8ffbc402ca2b6bcbf7cb01c4 · 2025-01-26 20:31:28 +0000 · Alex Cheema

Files touched

Diff

commit 82ef08601049be3f8ffbc402ca2b6bcbf7cb01c4
Author: Alex Cheema <alexcheema123@gmail.com>
Date:   Sun Jan 26 20:31:28 2025 +0000

    add deepseek-v3-3bit and deepseek-r1-3bit
---
 exo/models.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/exo/models.py b/exo/models.py
index 333e29c5..29019e91 100644
--- a/exo/models.py
+++ b/exo/models.py
@@ -89,7 +89,9 @@ model_cards = {
   "deepseek-coder-v2-lite": { "layers": 27, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-Coder-V2-Lite-Instruct-4bit-mlx", }, },
   "deepseek-coder-v2.5": { "layers": 60, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-V2.5-MLX-AQ4_1_64", }, },
   "deepseek-v3": { "layers": 61, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-V3-4bit", }, },
+  "deepseek-v3-3bit": { "layers": 61, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-V3-3bit", }, },
   "deepseek-r1": { "layers": 61, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-R1-4bit", }, },
+  "deepseek-r1-3bit": { "layers": 61, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-R1-3bit", }, },
   ### deepseek distills
   "deepseek-r1-distill-qwen-1.5b": { "layers": 28, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/deepseek-r1-distill-qwen-1.5b", }, },
   "deepseek-r1-distill-qwen-1.5b-3bit": { "layers": 28, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/DeepSeek-R1-Distill-Qwen-1.5B-3bit", }, },

← 55ea3669 fix post_init deepseek v3  ·  back to Exo  ·  add exception for mlx-community/DeepSeek-R1-3bit and mlx-com a3766f53 →