daniicruzz commited on
Commit
b12313d
ยท
verified ยท
1 Parent(s): 61b8271

Update model card: drop KV cache mentions

Browse files
Files changed (1) hide show
  1. README.md +2 -3
README.md CHANGED
@@ -58,7 +58,7 @@ The BF16 release applies **model compression techniques** to significantly reduc
58
  |-----------------------|-------------|
59
  | Base model | [nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16). **31.6B** total parameters, **3.6B** activated per forward pass (11.34% activation ratio). [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/). |
60
  | Pulsar (BF16 release) | [MultiverseComputingCAI/Pulsar-16B](https://huggingface.co/MultiverseComputingCAI/Pulsar-16B). **16.15B** total parameters, **3.1B** activated per forward pass (19.28% activation ratio) after CompactifAI compression. |
61
- | Pulsar (this model) | **16.15B** parameters quantized to **FP8** (weights & activations). KV cache stored in **FP8**. |
62
  | ๐Ÿ“ **Architecture** | Hybrid Mamba2-Transformer with MoE (same family as the base checkpoint). |
63
  | ๐Ÿ› ๏ธ **Tool calling** | Yes. Same tool-call structure and format as [Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16). See [Tool Calling](#tool-calling). |
64
  | ๐Ÿ—œ๏ธ **Compression** | CompactifAI (proprietary compression technology) |
@@ -74,7 +74,6 @@ This checkpoint is a **post-training-quantized (PTQ)** version of [Pulsar-16B](h
74
  | Field | Value |
75
  |---------------------|-------|
76
  | Quantization scheme | **FP8** โ€” weights & activations |
77
- | KV cache | **FP8** |
78
  | Producer | `modelopt` 0.42.0 |
79
  | Excluded modules | `lm_head`, all Mamba2 `conv1d` modules, and the attention/Mamba projections (`q/k/v/o_proj`, `in_proj`, `out_proj`) of selected layers (kept in higher precision). |
80
 
@@ -418,7 +417,7 @@ Trained mainly on English with added Spanish. No systematic evaluation for langu
418
  |--------------|--------------------- |
419
  | Model name | Pulsar 16B FP8 |
420
  | Based on | [MultiverseComputingCAI/Pulsar-16B](https://huggingface.co/MultiverseComputingCAI/Pulsar-16B) (originally [NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16)) |
421
- | Precision | FP8 weights & activations, FP8 KV cache |
422
  | Quantization | NVIDIA TensorRT Model Optimizer (`modelopt` 0.42.0) |
423
  | Version | v1.5.0 |
424
  | Release date | TBD |
 
58
  |-----------------------|-------------|
59
  | Base model | [nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16). **31.6B** total parameters, **3.6B** activated per forward pass (11.34% activation ratio). [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/). |
60
  | Pulsar (BF16 release) | [MultiverseComputingCAI/Pulsar-16B](https://huggingface.co/MultiverseComputingCAI/Pulsar-16B). **16.15B** total parameters, **3.1B** activated per forward pass (19.28% activation ratio) after CompactifAI compression. |
61
+ | Pulsar (this model) | **16.15B** parameters quantized to **FP8** (weights & activations). |
62
  | ๐Ÿ“ **Architecture** | Hybrid Mamba2-Transformer with MoE (same family as the base checkpoint). |
63
  | ๐Ÿ› ๏ธ **Tool calling** | Yes. Same tool-call structure and format as [Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16). See [Tool Calling](#tool-calling). |
64
  | ๐Ÿ—œ๏ธ **Compression** | CompactifAI (proprietary compression technology) |
 
74
  | Field | Value |
75
  |---------------------|-------|
76
  | Quantization scheme | **FP8** โ€” weights & activations |
 
77
  | Producer | `modelopt` 0.42.0 |
78
  | Excluded modules | `lm_head`, all Mamba2 `conv1d` modules, and the attention/Mamba projections (`q/k/v/o_proj`, `in_proj`, `out_proj`) of selected layers (kept in higher precision). |
79
 
 
417
  |--------------|--------------------- |
418
  | Model name | Pulsar 16B FP8 |
419
  | Based on | [MultiverseComputingCAI/Pulsar-16B](https://huggingface.co/MultiverseComputingCAI/Pulsar-16B) (originally [NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16)) |
420
+ | Precision | FP8 weights & activations |
421
  | Quantization | NVIDIA TensorRT Model Optimizer (`modelopt` 0.42.0) |
422
  | Version | v1.5.0 |
423
  | Release date | TBD |