Preserve Needle bfloat16 runtime behavior

README.md

@@ -1,7 +1,6 @@
 ---
 license: mit
 library_name: mlx
-base_model: Cactus-Compute/needle
 tags:
   - mlx
   - function-calling
@@ -12,10 +11,14 @@ tags:
 
 # Needle MLX
 
-An MLX safetensors conversion of
+An MLX safetensors format conversion of
 [Cactus-Compute/needle](https://huggingface.co/Cactus-Compute/needle), a
 26M-parameter encoder-decoder function-calling model.
 
+This is not a fine-tune or a quantization. The published checkpoint is loaded
+using Needle's declared `bfloat16` inference dtype, then stored as MLX
+safetensors so it preserves the upstream runtime's executable values.
+
 ## What is included
 
 - `model.safetensors`: 31 converted tensors, 26,315,421 parameters.
@@ -40,15 +43,16 @@ dedicated adapter to the serving runtime.
 ## Conversion and verification
 
 The published `needle.pkl` checkpoint was read with a restricted NumPy-only
-pickle loader, converted directly to MLX safetensors, and checked
-tensor-for-tensor against the source.
+pickle loader, converted to MLX `bfloat16` safetensors, and checked
+tensor-for-tensor after the same dtype cast used by Needle's upstream runtime.
 
 - Source SHA-256:
   `40a32e91d1d4197bf15ba559b74f6727c342dc8746918742fc7d8e2c1f18df40`
 - Converted `model.safetensors` SHA-256:
   `7b9d5f0d6ddeb7fbb20f4e45f3f616919357e5d08b5778859fdb762a33d60dae`
-- Verification: 31/31 tensors equal the source; an MLX encoder-decoder smoke
-  pass produced logits with shape `(1, 2, 8192)`.
+- Verification: 31/31 tensors equal the source after Needle's `bfloat16`
+  runtime cast; an MLX encoder-decoder smoke pass produced logits with shape
+  `(1, 2, 8192)` and generated the upstream weather tool-call example.
 
 The converter source is available at
 [seeker-cyber-maker/needle-mlx-depicklinator](https://github.com/seeker-cyber-maker/needle-mlx-depicklinator).

config.json

@@ -22,7 +22,8 @@
     "NeedleModel"
   ],
   "format": "mlx-safetensors",
-  "format_version": 1,
+  "format_version": 2,
+  "tensor_dtype": "bfloat16",
   "eos_token_id": 1,
   "source_checkpoint": "needle.pkl",
   "source_sha256": "40a32e91d1d4197bf15ba559b74f6727c342dc8746918742fc7d8e2c1f18df40"

manifest.json

@@ -1,8 +1,9 @@
 {
   "format": "needle-mlx",
-  "format_version": 1,
+  "format_version": 2,
   "tensor_count": 31,
   "parameter_count": 26315421,
+  "tensor_dtype": "bfloat16",
   "source_sha256": "40a32e91d1d4197bf15ba559b74f6727c342dc8746918742fc7d8e2c1f18df40",
   "source_checkpoint": "needle.pkl"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b9d5f0d6ddeb7fbb20f4e45f3f616919357e5d08b5778859fdb762a33d60dae
-size 52634442
+oid sha256:8491dba478deb60e46e437714911e1db729a6463f6289505f99af472b6aac7bc
+size 52634473