Instructions to use PakNin/Reuse-Trained-R2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PakNin/Reuse-Trained-R2 with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("microsoft/Phi-mini-MoE-instruct")
model = PeftModel.from_pretrained(base_model, "PakNin/Reuse-Trained-R2")

Transformers

How to use PakNin/Reuse-Trained-R2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="PakNin/Reuse-Trained-R2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("PakNin/Reuse-Trained-R2", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use PakNin/Reuse-Trained-R2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PakNin/Reuse-Trained-R2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PakNin/Reuse-Trained-R2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PakNin/Reuse-Trained-R2

SGLang

How to use PakNin/Reuse-Trained-R2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "PakNin/Reuse-Trained-R2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PakNin/Reuse-Trained-R2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "PakNin/Reuse-Trained-R2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PakNin/Reuse-Trained-R2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use PakNin/Reuse-Trained-R2 with Docker Model Runner:
```
docker model run hf.co/PakNin/Reuse-Trained-R2
```

PakNin commited on 2 days ago

Commit

a0ece80

verified ·

1 Parent(s): ee0db66

Upload folder using huggingface_hub

Browse files

Files changed (22) hide show

README.md +207 -0
adapter_config.json +46 -0
adapter_model.safetensors +3 -0
chat_template.jinja +4 -0
logs/rexmoe_training_0204_041628.log +365 -0
merged/chat_template.jinja +4 -0
merged/config.json +41 -0
merged/configuration_slimmoe.py +248 -0
merged/generation_config.json +11 -0
merged/model-00001-of-00004.safetensors +3 -0
merged/model-00002-of-00004.safetensors +3 -0
merged/model-00003-of-00004.safetensors +3 -0
merged/model-00004-of-00004.safetensors +3 -0
merged/model.safetensors.index.json +0 -0
merged/modeling_slimmoe.py +1800 -0
merged/special_tokens_map.json +30 -0
merged/tokenizer.json +0 -0
merged/tokenizer_config.json +131 -0
rexmoe_routers.pt +3 -0
special_tokens_map.json +30 -0
tokenizer.json +0 -0
tokenizer_config.json +131 -0

README.md ADDED Viewed

	@@ -0,0 +1,207 @@

+---
+base_model: microsoft/Phi-mini-MoE-instruct
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:microsoft/Phi-mini-MoE-instruct
+- lora
+- transformers
+---
+# Model Card for Model ID
+<!-- Provide a quick summary of what the model is/does. -->
+## Model Details
+### Model Description
+<!-- Provide a longer summary of what this model is. -->
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+### Model Sources [optional]
+<!-- Provide the basic links for the model. -->
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+## Uses
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+### Direct Use
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+[More Information Needed]
+### Downstream Use [optional]
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+[More Information Needed]
+### Out-of-Scope Use
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+[More Information Needed]
+## Bias, Risks, and Limitations
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+[More Information Needed]
+### Recommendations
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+## How to Get Started with the Model
+Use the code below to get started with the model.
+[More Information Needed]
+## Training Details
+### Training Data
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+[More Information Needed]
+### Training Procedure
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+#### Preprocessing [optional]
+[More Information Needed]
+#### Training Hyperparameters
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+#### Speeds, Sizes, Times [optional]
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+[More Information Needed]
+## Evaluation
+<!-- This section describes the evaluation protocols and provides the results. -->
+### Testing Data, Factors & Metrics
+#### Testing Data
+<!-- This should link to a Dataset Card if possible. -->
+[More Information Needed]
+#### Factors
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+[More Information Needed]
+#### Metrics
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+[More Information Needed]
+### Results
+[More Information Needed]
+#### Summary
+## Model Examination [optional]
+<!-- Relevant interpretability work for the model goes here -->
+[More Information Needed]
+## Environmental Impact
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+## Technical Specifications [optional]
+### Model Architecture and Objective
+[More Information Needed]
+### Compute Infrastructure
+[More Information Needed]
+#### Hardware
+[More Information Needed]
+#### Software
+[More Information Needed]
+## Citation [optional]
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+**BibTeX:**
+[More Information Needed]
+**APA:**
+[More Information Needed]
+## Glossary [optional]
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+[More Information Needed]
+## More Information [optional]
+[More Information Needed]
+## Model Card Authors [optional]
+[More Information Needed]
+## Model Card Contact
+[More Information Needed]
+### Framework versions
+- PEFT 0.18.1

adapter_config.json ADDED Viewed

	@@ -0,0 +1,46 @@

+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "microsoft/Phi-mini-MoE-instruct",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "w2",
+    "o_proj",
+    "w1",
+    "w3",
+    "q_proj",
+    "k_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:b64460c5d5ff01c99275896e759087525336b6a2ff5efa9a5eece0cc68562031
+size 552044120

chat_template.jinja ADDED Viewed

	@@ -0,0 +1,4 @@

+{% for message in messages %}{{'<|' + message['role'] + '|>' + '
+' + message['content'] + '<|end|>
+' }}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>
+' }}{% else %}{{ eos_token }}{% endif %}

logs/rexmoe_training_0204_041628.log ADDED Viewed

	@@ -0,0 +1,365 @@

+2026-04-02 04:16:28 - ReXMoE - INFO - ================================================================================
+2026-04-02 04:16:28 - ReXMoE - INFO - ReXMoE Training Log - 0204_041628
+2026-04-02 04:16:28 - ReXMoE - INFO - Log file: ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2/logs/rexmoe_training_0204_041628.log
+2026-04-02 04:16:28 - ReXMoE - INFO - ================================================================================
+2026-04-02 04:16:28 - ReXMoE - INFO - ================================================================================
+2026-04-02 04:16:28 - ReXMoE - INFO - ReXMoE Cross-Layer Expert Reuse Training
+2026-04-02 04:16:28 - ReXMoE - INFO - ================================================================================
+2026-04-02 04:16:28 - ReXMoE - INFO - MET enabled: False
+2026-04-02 04:16:28 - ReXMoE - INFO -
+Configuration:
+  Model: microsoft/Phi-mini-MoE-instruct
+  Dataset: ../dataset/alpaca_data_cleaned.json
+  Dataset mode: IF_2
+  Reuse Scale (R): 2
+  Prune Ratio (MET): N/A
+  Epochs: 2
+  Num of samples: 20000
+  Batch Size: 2
+  Sequence Length: 1024
+  Learning Rate: 5e-05
+  PSR Enabled: True
+  LR Scheduler: True
+  Save Path: ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2
+  Gradient Checkpointing: False
+  LoRA Rank: 16 (Full LoRA: True)
+  LoRA Alpha: 32
+  MET Enabled: False (Mask Ratio: 0.1, Warmup: 0.5)
+  Log File: ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2/logs/rexmoe_training_0204_041628.log
+  Aux loss weight: 0.05
+2026-04-02 04:16:28 - ReXMoE - INFO - 💻 Using device: cuda)
+2026-04-02 04:16:28 - ReXMoE - INFO - GPU: NVIDIA RTX A6000, Memory: 47.53 GB
+2026-04-02 04:16:37 - ReXMoE - INFO - [5/7] Setting up optimizer and dataset...
+2026-04-02 04:16:37 - ReXMoE - INFO - Using 8-bit AdamW optimizer
+2026-04-02 04:16:37 - ReXMoE - INFO - LR Scheduler: CosineAnnealingLR (5e-05 → 5e-06)
+2026-04-02 04:16:45 - ReXMoE - INFO -
+  First batch statistics:
+2026-04-02 04:16:45 - ReXMoE - INFO -     LM Loss: 1.0013
+2026-04-02 04:16:45 - ReXMoE - INFO -     Aux Loss: 0.108887
+2026-04-02 04:16:45 - ReXMoE - INFO -     Total Loss: 1.1102
+2026-04-02 04:16:45 - ReXMoE - INFO -     Current R: 2
+2026-04-02 04:16:45 - ReXMoE - INFO -     Active experts per layer: 32
+2026-04-02 04:16:45 - ReXMoE - INFO -     Gradient norm: 1.0000
+2026-04-02 04:16:45 - ReXMoE - INFO -
+2026-04-02 04:21:32 - ReXMoE - INFO -   [100/10000] loss=2.2108 aux=0.046143 R=2
+2026-04-02 04:26:17 - ReXMoE - INFO -   [200/10000] loss=0.7772 aux=0.018188 R=2
+2026-04-02 04:31:00 - ReXMoE - INFO -   [300/10000] loss=0.9490 aux=0.021973 R=2
+2026-04-02 04:35:41 - ReXMoE - INFO -   [400/10000] loss=0.6892 aux=0.051025 R=2
+2026-04-02 04:40:24 - ReXMoE - INFO -   [500/10000] loss=1.5216 aux=0.014099 R=2
+2026-04-02 04:45:08 - ReXMoE - INFO -   [600/10000] loss=1.6429 aux=0.006897 R=2
+2026-04-02 04:49:52 - ReXMoE - INFO -   [700/10000] loss=1.0995 aux=0.008118 R=2
+2026-04-02 04:54:33 - ReXMoE - INFO -   [800/10000] loss=1.4248 aux=0.006561 R=2
+2026-04-02 04:59:14 - ReXMoE - INFO -   [900/10000] loss=1.7497 aux=0.008850 R=2
+2026-04-02 05:03:55 - ReXMoE - INFO - Warmup completed at step 1000. Enabling FULL QLoRA with r = 16 and alpha = 32 on experts and updating optimizer...
+2026-04-02 05:04:09 - ReXMoE - INFO - Trainable params (routers + LoRA): 142082048 (1.8245%)
+2026-04-02 05:04:09 - ReXMoE - INFO - Sample trainable params after QLoRA: ['base_model.model.model.layers.0.self_attn.q_proj.lora_A.default.weight', 'base_model.model.model.layers.0.self_attn.q_proj.lora_B.default.weight', 'base_model.model.model.layers.0.self_attn.k_proj.lora_A.default.weight', 'base_model.model.model.layers.0.self_attn.k_proj.lora_B.default.weight', 'base_model.model.model.layers.0.self_attn.v_proj.lora_A.default.weight', 'base_model.model.model.layers.0.self_attn.v_proj.lora_B.default.weight', 'base_model.model.model.layers.0.self_attn.o_proj.lora_A.default.weight', 'base_model.model.model.layers.0.self_attn.o_proj.lora_B.default.weight', 'base_model.model.model.layers.0.block_sparse_moe.gate.weight', 'base_model.model.model.layers.0.block_sparse_moe.experts.0.w1.lora_A.default.weight']
+2026-04-02 05:04:16 - ReXMoE - INFO -   [1000/10000] loss=0.7002 aux=0.005341 R=2
+2026-04-02 05:13:07 - ReXMoE - INFO -   [1100/10000] loss=0.6426 aux=0.007233 R=2
+2026-04-02 05:21:43 - ReXMoE - INFO -   [1200/10000] loss=0.5568 aux=0.007019 R=2
+2026-04-02 05:30:24 - ReXMoE - INFO -   [1300/10000] loss=0.8270 aux=0.006531 R=2
+2026-04-02 05:39:34 - ReXMoE - INFO -   [1400/10000] loss=0.8769 aux=0.005249 R=2
+2026-04-02 05:48:47 - ReXMoE - INFO -   [1500/10000] loss=0.6700 aux=0.010498 R=2
+2026-04-02 05:57:41 - ReXMoE - INFO -   [1600/10000] loss=0.6573 aux=0.004242 R=2
+2026-04-02 06:06:26 - ReXMoE - INFO -   [1700/10000] loss=0.8331 aux=0.002289 R=2
+2026-04-02 06:15:13 - ReXMoE - INFO -   [1800/10000] loss=1.0478 aux=0.010437 R=2
+2026-04-02 06:24:01 - ReXMoE - INFO -   [1900/10000] loss=1.7566 aux=0.002350 R=2
+2026-04-02 06:32:51 - ReXMoE - INFO -   [2000/10000] loss=0.9031 aux=0.001572 R=2
+2026-04-02 06:41:39 - ReXMoE - INFO -   [2100/10000] loss=0.8975 aux=0.005035 R=2
+2026-04-02 06:50:29 - ReXMoE - INFO -   [2200/10000] loss=1.4439 aux=0.021606 R=2
+2026-04-02 06:59:22 - ReXMoE - INFO -   [2300/10000] loss=1.1146 aux=0.003876 R=2
+2026-04-02 07:08:13 - ReXMoE - INFO -   [2400/10000] loss=1.2936 aux=0.003204 R=2
+2026-04-02 07:17:11 - ReXMoE - INFO -   [2500/10000] loss=0.5071 aux=0.002731 R=2
+2026-04-02 07:26:02 - ReXMoE - INFO -   [2600/10000] loss=0.6228 aux=0.004730 R=2
+2026-04-02 07:34:42 - ReXMoE - INFO -   [2700/10000] loss=0.6085 aux=0.099609 R=2
+2026-04-02 07:43:17 - ReXMoE - INFO -   [2800/10000] loss=0.6629 aux=0.003265 R=2
+2026-04-02 07:51:52 - ReXMoE - INFO -   [2900/10000] loss=0.4027 aux=0.002472 R=2
+2026-04-02 08:00:24 - ReXMoE - INFO -   [3000/10000] loss=0.5227 aux=0.002045 R=2
+2026-04-02 08:09:03 - ReXMoE - INFO -   [3100/10000] loss=1.5149 aux=0.001709 R=2
+2026-04-02 08:17:32 - ReXMoE - INFO -   [3200/10000] loss=0.8226 aux=0.006104 R=2
+2026-04-02 08:26:01 - ReXMoE - INFO -   [3300/10000] loss=1.2210 aux=0.001472 R=2
+2026-04-02 08:34:30 - ReXMoE - INFO -   [3400/10000] loss=0.8192 aux=0.008179 R=2
+2026-04-02 08:43:04 - ReXMoE - INFO -   [3500/10000] loss=1.1357 aux=0.001709 R=2
+2026-04-02 08:51:39 - ReXMoE - INFO -   [3600/10000] loss=2.2026 aux=0.003387 R=2
+2026-04-02 09:00:07 - ReXMoE - INFO -   [3700/10000] loss=1.1631 aux=0.010376 R=2
+2026-04-02 09:08:41 - ReXMoE - INFO -   [3800/10000] loss=1.9066 aux=0.001495 R=2
+2026-04-02 09:17:11 - ReXMoE - INFO -   [3900/10000] loss=0.4675 aux=0.002136 R=2
+2026-04-02 09:25:43 - ReXMoE - INFO -   [4000/10000] loss=0.6906 aux=0.002838 R=2
+2026-04-02 09:34:13 - ReXMoE - INFO -   [4100/10000] loss=1.1627 aux=0.003693 R=2
+2026-04-02 09:42:49 - ReXMoE - INFO -   [4200/10000] loss=1.0517 aux=0.001595 R=2
+2026-04-02 09:51:25 - ReXMoE - INFO -   [4300/10000] loss=0.4299 aux=0.003799 R=2
+2026-04-02 09:59:56 - ReXMoE - INFO -   [4400/10000] loss=0.6066 aux=0.002121 R=2
+2026-04-02 10:08:25 - ReXMoE - INFO -   [4500/10000] loss=1.0687 aux=0.027466 R=2
+2026-04-02 10:16:55 - ReXMoE - INFO -   [4600/10000] loss=0.6690 aux=0.004272 R=2
+2026-04-02 10:25:25 - ReXMoE - INFO -   [4700/10000] loss=0.7443 aux=0.001610 R=2
+2026-04-02 10:33:55 - ReXMoE - INFO -   [4800/10000] loss=0.5073 aux=0.001488 R=2
+2026-04-02 10:42:24 - ReXMoE - INFO -   [4900/10000] loss=0.8397 aux=0.004517 R=2
+2026-04-02 10:50:56 - ReXMoE - INFO -   [5000/10000] loss=0.4906 aux=0.001007 R=2
+2026-04-02 10:59:27 - ReXMoE - INFO -   [5100/10000] loss=0.6861 aux=0.002716 R=2
+2026-04-02 11:07:57 - ReXMoE - INFO -   [5200/10000] loss=0.4963 aux=0.002823 R=2
+2026-04-02 11:16:24 - ReXMoE - INFO -   [5300/10000] loss=0.7556 aux=0.002533 R=2
+2026-04-02 11:24:50 - ReXMoE - INFO -   [5400/10000] loss=0.6053 aux=0.020874 R=2
+2026-04-02 11:33:16 - ReXMoE - INFO -   [5500/10000] loss=2.0345 aux=0.000778 R=2
+2026-04-02 11:41:41 - ReXMoE - INFO -   [5600/10000] loss=0.7234 aux=0.025269 R=2
+2026-04-02 11:50:07 - ReXMoE - INFO -   [5700/10000] loss=0.3542 aux=0.000467 R=2
+2026-04-02 11:58:34 - ReXMoE - INFO -   [5800/10000] loss=0.8516 aux=0.002274 R=2
+2026-04-02 12:07:01 - ReXMoE - INFO -   [5900/10000] loss=0.5901 aux=0.001198 R=2
+2026-04-02 12:15:26 - ReXMoE - INFO -   [6000/10000] loss=1.3560 aux=0.001457 R=2
+2026-04-02 12:23:51 - ReXMoE - INFO -   [6100/10000] loss=1.2492 aux=0.000854 R=2
+2026-04-02 12:32:18 - ReXMoE - INFO -   [6200/10000] loss=0.3890 aux=0.001007 R=2
+2026-04-02 12:40:43 - ReXMoE - INFO -   [6300/10000] loss=0.5883 aux=0.003555 R=2
+2026-04-02 12:49:08 - ReXMoE - INFO -   [6400/10000] loss=0.5396 aux=0.001968 R=2
+2026-04-02 12:57:35 - ReXMoE - INFO -   [6500/10000] loss=1.0615 aux=0.000652 R=2
+2026-04-02 13:06:04 - ReXMoE - INFO -   [6600/10000] loss=0.8583 aux=0.003906 R=2
+2026-04-02 13:14:33 - ReXMoE - INFO -   [6700/10000] loss=0.4472 aux=0.001663 R=2
+2026-04-02 13:23:03 - ReXMoE - INFO -   [6800/10000] loss=1.3548 aux=0.001114 R=2
+2026-04-02 13:31:35 - ReXMoE - INFO -   [6900/10000] loss=1.0129 aux=0.001389 R=2
+2026-04-02 13:40:04 - ReXMoE - INFO -   [7000/10000] loss=1.3876 aux=0.000656 R=2
+2026-04-02 13:48:35 - ReXMoE - INFO -   [7100/10000] loss=1.0568 aux=0.002686 R=2
+2026-04-02 13:57:05 - ReXMoE - INFO -   [7200/10000] loss=1.7856 aux=0.001724 R=2
+2026-04-02 14:05:34 - ReXMoE - INFO -   [7300/10000] loss=0.9223 aux=0.000748 R=2
+2026-04-02 14:14:03 - ReXMoE - INFO -   [7400/10000] loss=0.3757 aux=0.021851 R=2
+2026-04-02 14:22:34 - ReXMoE - INFO -   [7500/10000] loss=0.8600 aux=0.010559 R=2
+2026-04-02 14:31:04 - ReXMoE - INFO -   [7600/10000] loss=0.8164 aux=0.002304 R=2
+2026-04-02 14:39:35 - ReXMoE - INFO -   [7700/10000] loss=1.2134 aux=0.001442 R=2
+2026-04-02 14:48:04 - ReXMoE - INFO -   [7800/10000] loss=0.8161 aux=0.005493 R=2
+2026-04-02 14:56:35 - ReXMoE - INFO -   [7900/10000] loss=1.6585 aux=0.001915 R=2
+2026-04-02 15:05:04 - ReXMoE - INFO -   [8000/10000] loss=1.2704 aux=0.000740 R=2
+2026-04-02 15:13:35 - ReXMoE - INFO -   [8100/10000] loss=0.5500 aux=0.001045 R=2
+2026-04-02 15:22:06 - ReXMoE - INFO -   [8200/10000] loss=1.5001 aux=0.001602 R=2
+2026-04-02 15:30:37 - ReXMoE - INFO -   [8300/10000] loss=0.5917 aux=0.000664 R=2
+2026-04-02 15:39:07 - ReXMoE - INFO -   [8400/10000] loss=0.1938 aux=0.001656 R=2
+2026-04-02 15:47:35 - ReXMoE - INFO -   [8500/10000] loss=0.6856 aux=0.006897 R=2
+2026-04-02 15:56:03 - ReXMoE - INFO -   [8600/10000] loss=1.3575 aux=0.001816 R=2
+2026-04-02 16:04:31 - ReXMoE - INFO -   [8700/10000] loss=0.4273 aux=0.003571 R=2
+2026-04-02 16:13:00 - ReXMoE - INFO -   [8800/10000] loss=0.7848 aux=0.001518 R=2
+2026-04-02 16:21:28 - ReXMoE - INFO -   [8900/10000] loss=0.4080 aux=0.001038 R=2
+2026-04-02 16:29:58 - ReXMoE - INFO -   [9000/10000] loss=0.5492 aux=0.005341 R=2
+2026-04-02 16:38:27 - ReXMoE - INFO -   [9100/10000] loss=1.4560 aux=0.000690 R=2
+2026-04-02 16:46:57 - ReXMoE - INFO -   [9200/10000] loss=0.8392 aux=0.000916 R=2
+2026-04-02 16:55:25 - ReXMoE - INFO -   [9300/10000] loss=1.3378 aux=0.002823 R=2
+2026-04-02 17:03:56 - ReXMoE - INFO -   [9400/10000] loss=1.1250 aux=0.001312 R=2
+2026-04-02 17:12:27 - ReXMoE - INFO -   [9500/10000] loss=0.9452 aux=0.000587 R=2
+2026-04-02 17:20:56 - ReXMoE - INFO -   [9600/10000] loss=0.3425 aux=0.015869 R=2
+2026-04-02 17:29:25 - ReXMoE - INFO -   [9700/10000] loss=1.1245 aux=0.001831 R=2
+2026-04-02 17:37:52 - ReXMoE - INFO -   [9800/10000] loss=0.8069 aux=0.001640 R=2
+2026-04-02 17:46:19 - ReXMoE - INFO -   [9900/10000] loss=1.7475 aux=0.002563 R=2
+2026-04-02 17:54:44 - ReXMoE - INFO -
+[Step 10000/20000] Running evaluation at eval_steps...
+2026-04-02 17:54:44 - ReXMoE - INFO -
+Evaluating model with 3 sample prompts...
+2026-04-02 17:54:47 - ReXMoE - INFO -
+--- Prompt 1/3 ---
+2026-04-02 17:54:47 - ReXMoE - INFO - Instruction: What is the capital of France?
+2026-04-02 17:54:47 - ReXMoE - INFO - Input: None
+2026-04-02 17:54:47 - ReXMoE - INFO - Generated completion (len 9): The capital of France is Paris.
+2026-04-02 17:54:49 - ReXMoE - INFO -
+--- Prompt 2/3 ---
+2026-04-02 17:54:49 - ReXMoE - INFO - Instruction: High-pressure systems stop air from rising into the colder regions of the atmosphere where water can condense. What will most likely result if a high-pressure system remains in an area for a long period of time?
+A. fog
+B. rain
+C. drought
+D. tornado
+Answer:
+2026-04-02 17:54:49 - ReXMoE - INFO - Input: None
+2026-04-02 17:54:49 - ReXMoE - INFO - Generated completion (len 6): C. drought
+2026-04-02 17:54:50 - ReXMoE - INFO -
+--- Prompt 3/3 ---
+2026-04-02 17:54:50 - ReXMoE - INFO - Instruction: Given the fact: predators eat prey
+Question: Predators eat
+A. lions
+B. humans
+C. bunnies
+D. grass
+Answer:
+2026-04-02 17:54:50 - ReXMoE - INFO - Input: None
+2026-04-02 17:54:50 - ReXMoE - INFO - Generated completion (len 7): C. bunnies
+2026-04-02 17:54:50 - ReXMoE - INFO - Evaluation of all 3 prompts complete.
+2026-04-02 17:54:50 - ReXMoE - INFO -
+[Step 10000] Analyzing routing patterns at eval_steps...
+2026-04-02 17:55:07 - ReXMoE - INFO -
+Analyzing ACTUAL routing patterns from 10 batches (7,610 tokens)
+2026-04-02 17:55:07 - ReXMoE - INFO - Current reuse scale: R=2
+2026-04-02 17:55:07 - ReXMoE - INFO -
+[IG-MET Pruning Report]:
+2026-04-02 17:55:07 - ReXMoE - INFO - Global: 0/0 UNIQUE experts pruned (0.0%) | threshold=-1.000000
+2026-04-02 17:55:07 - ReXMoE - INFO - Cross-Layer Routing Distribution (ACTUAL selections):
+2026-04-02 17:55:07 - ReXMoE - INFO -   Same layer (i):              639,745 ( 48.8%)
+2026-04-02 17:55:07 - ReXMoE - INFO -   Previous layer (i-1):         18,093 (  1.4%)
+2026-04-02 17:55:07 - ReXMoE - INFO -   Next layer (i+1):            652,882 ( 49.8%)
+2026-04-02 17:55:07 - ReXMoE - INFO - Sample Layer-Specific Routing Patterns:
+2026-04-02 17:55:07 - ReXMoE - INFO -
+  Layer 8:
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert 14 from layer  9 (  L9):    5,114 times ( 67.2%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  2 from layer  9 (  L9):    3,743 times ( 49.2%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  7 from layer  9 (  L9):    3,668 times ( 48.2%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert 11 from layer  9 (  L9):    2,931 times ( 38.5%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert 13 from layer  9 (  L9):    2,786 times ( 36.6%)
+2026-04-02 17:55:07 - ReXMoE - INFO -
+  Layer 16:
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  8 from layer 17 ( L17):    6,835 times ( 89.8%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  8 from layer 16 (same):    5,658 times ( 74.3%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert 10 from layer 17 ( L17):    3,268 times ( 42.9%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  1 from layer 17 ( L17):    2,518 times ( 33.1%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  9 from layer 16 (same):    1,967 times ( 25.8%)
+2026-04-02 17:55:07 - ReXMoE - INFO -
+  Layer 24:
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  8 from layer 25 ( L25):    6,353 times ( 83.5%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  8 from layer 24 (same):    5,423 times ( 71.3%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  9 from layer 25 ( L25):    3,424 times ( 45.0%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert 10 from layer 24 (same):    3,301 times ( 43.4%)
+2026-04-02 17:55:07 - ReXMoE - INFO -     Expert  4 from layer 24 (same):    1,978 times ( 26.0%)
+2026-04-02 17:55:07 - ReXMoE - INFO - ✅ Cross-layer expert reuse detected: 51.2% of routing uses adjacent layers
+2026-04-02 17:55:07 - ReXMoE - INFO -
+[Step 10000] Saving checkpoint at eval_steps to ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2...
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.0.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.0.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.1.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.1.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.2.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.2.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.3.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.3.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.4.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.4.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.5.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.5.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.6.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.6.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.7.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.7.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.8.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.8.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.9.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.9.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.10.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.10.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.11.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.11.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.12.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.12.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.13.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.13.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.14.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.14.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.15.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.15.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.16.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.16.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.17.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.17.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.18.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.18.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.19.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.19.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.20.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.20.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.21.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.21.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.22.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.22.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.23.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.23.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.24.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.24.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.25.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.25.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.26.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.26.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.27.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.27.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.28.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.28.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.29.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.29.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.30.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.30.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.31.block_sparse_moe.router.ema_utilization with shape torch.Size([32]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - Saving buffer base_model.model.model.layers.31.block_sparse_moe.router.mask_threshold with shape torch.Size([]) for pruning evaluation
+2026-04-02 17:55:07 - ReXMoE - INFO - ✓ Saved trained router weights: 96 parameters
+2026-04-02 17:55:07 - ReXMoE - INFO -   File: ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2/rexmoe_routers.pt
+2026-04-02 17:55:07 - ReXMoE - INFO -   Size: 8.03 MB
+2026-04-02 17:55:07 - ReXMoE - INFO -
+Also saving full model with ReXMoE architecture...
+2026-04-02 17:55:08 - ReXMoE - INFO -
+Merging LoRA adapters into base weights and saving to: ./0204_041628_10_rexmoe_natural_phi_mini_moe_R2/merged
+2026-04-02 17:55:25 - ReXMoE - INFO - ✓ Saved merged full model (base+routers+LoRA) for one-step loading
+2026-04-02 17:55:25 - ReXMoE - INFO -
+============================================================
+2026-04-02 17:55:25 - ReXMoE - INFO - Epoch 1 Summary:
+2026-04-02 17:55:25 - ReXMoE - INFO -   Average LM Loss: nan
+2026-04-02 17:55:25 - ReXMoE - INFO -   Average Aux Loss: 0.007334
+2026-04-02 17:55:25 - ReXMoE - INFO -   Average Total Loss: nan
+2026-04-02 17:55:25 - ReXMoE - INFO -   Final R: 2
+2026-04-02 17:55:25 - ReXMoE - INFO -
+Evaluating model with 3 sample prompts...
+2026-04-02 17:55:26 - ReXMoE - INFO -
+--- Prompt 1/3 ---
+2026-04-02 17:55:26 - ReXMoE - INFO - Instruction: What is the capital of France?
+2026-04-02 17:55:26 - ReXMoE - INFO - Input: None
+2026-04-02 17:55:26 - ReXMoE - INFO - Generated completion (len 9): The capital of France is Paris.
+2026-04-02 17:55:28 - ReXMoE - INFO -
+--- Prompt 2/3 ---
+2026-04-02 17:55:28 - ReXMoE - INFO - Instruction: High-pressure systems stop air from rising into the colder regions of the atmosphere where water can condense. What will most likely result if a high-pressure system remains in an area for a long period of time?
+A. fog
+B. rain
+C. drought
+D. tornado
+Answer:
+2026-04-02 17:55:28 - ReXMoE - INFO - Input: None
+2026-04-02 17:55:28 - ReXMoE - INFO - Generated completion (len 6): C. drought
+2026-04-02 17:55:29 - ReXMoE - INFO -
+--- Prompt 3/3 ---
+2026-04-02 17:55:29 - ReXMoE - INFO - Instruction: Given the fact: predators eat prey
+Question: Predators eat
+A. lions
+B. humans
+C. bunnies
+D. grass
+Answer:
+2026-04-02 17:55:29 - ReXMoE - INFO - Input: None
+2026-04-02 17:55:29 - ReXMoE - INFO - Generated completion (len 7): C. bunnies
+2026-04-02 17:55:29 - ReXMoE - INFO - Evaluation of all 3 prompts complete.
+2026-04-02 17:55:29 - ReXMoE - INFO -
+  📊 Convergence Metrics:
+2026-04-02 17:55:29 - ReXMoE - INFO - Convergence Metrics:
+2026-04-02 17:55:29 - ReXMoE - INFO -     Avg Router Grad Norm: 0.100342
+2026-04-02 17:55:29 - ReXMoE - INFO -   Current Learning Rate: 5.00e-05
+2026-04-02 17:55:29 - ReXMoE - INFO - ℹ️  Collecting convergence data (need 5 epochs minimum)...
+2026-04-02 17:55:29 - ReXMoE - INFO - Routing Pattern Analysis (Epoch 1):
+2026-04-02 17:55:39 - ReXMoE - INFO -
+Analyzing ACTUAL routing patterns from 10 batches (7,965 tokens)
+2026-04-02 17:55:39 - ReXMoE - INFO - Current reuse scale: R=2
+2026-04-02 17:55:39 - ReXMoE - INFO -
+[IG-MET Pruning Report]:
+2026-04-02 17:55:39 - ReXMoE - INFO - Global: 0/0 UNIQUE experts pruned (0.0%) | threshold=-1.000000
+2026-04-02 17:55:39 - ReXMoE - INFO - Cross-Layer Routing Distribution (ACTUAL selections):
+2026-04-02 17:55:39 - ReXMoE - INFO -   Same layer (i):              636,430 ( 48.6%)
+2026-04-02 17:55:39 - ReXMoE - INFO -   Previous layer (i-1):         19,584 (  1.5%)
+2026-04-02 17:55:39 - ReXMoE - INFO -   Next layer (i+1):            654,706 ( 50.0%)
+2026-04-02 17:55:39 - ReXMoE - INFO - Sample Layer-Specific Routing Patterns:
+2026-04-02 17:55:39 - ReXMoE - INFO -
+  Layer 8:
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 14 from layer  9 (  L9):    4,872 times ( 61.2%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 13 from layer  9 (  L9):    3,980 times ( 50.0%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  7 from layer  9 (  L9):    3,882 times ( 48.7%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  2 from layer  9 (  L9):    3,156 times ( 39.6%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 11 from layer  9 (  L9):    3,019 times ( 37.9%)
+2026-04-02 17:55:39 - ReXMoE - INFO -
+  Layer 16:
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  8 from layer 17 ( L17):    6,786 times ( 85.2%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  8 from layer 16 (same):    5,551 times ( 69.7%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 10 from layer 17 ( L17):    3,275 times ( 41.1%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 15 from layer 17 ( L17):    2,236 times ( 28.1%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 10 from layer 16 (same):    2,116 times ( 26.6%)
+2026-04-02 17:55:39 - ReXMoE - INFO -
+  Layer 24:
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  8 from layer 25 ( L25):    6,017 times ( 75.5%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  8 from layer 24 (same):    4,472 times ( 56.1%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  9 from layer 25 ( L25):    4,132 times ( 51.9%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert 10 from layer 24 (same):    3,410 times ( 42.8%)
+2026-04-02 17:55:39 - ReXMoE - INFO -     Expert  4 from layer 24 (same):    2,419 times ( 30.4%)
+2026-04-02 17:55:39 - ReXMoE - INFO - ✅ Cross-layer expert reuse detected: 51.4% of routing uses adjacent layers
+2026-04-02 17:55:39 - ReXMoE - INFO - LR stepped to: 5.00e-05
+2026-04-02 18:00:28 - ReXMoE - INFO -   [100/10000] loss=0.5745 aux=0.001076 R=2

merged/chat_template.jinja ADDED Viewed

	@@ -0,0 +1,4 @@

+{% for message in messages %}{{'<|' + message['role'] + '|>' + '
+' + message['content'] + '<|end|>
+' }}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>
+' }}{% else %}{{ eos_token }}{% endif %}

merged/config.json ADDED Viewed

	@@ -0,0 +1,41 @@

+{
+  "architectures": [
+    "PhimoeForCausalLM"
+  ],
+  "attention_bias": true,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_slimmoe.PhiMoEConfig",
+    "AutoModelForCausalLM": "modeling_slimmoe.PhiMoEForCausalLM"
+  },
+  "bos_token_id": 1,
+  "dtype": "bfloat16",
+  "eos_token_id": 32000,
+  "expert_dropout": 0.0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "input_jitter_noise": 0.01,
+  "intermediate_size": 960,
+  "lm_head_bias": true,
+  "max_position_embeddings": 4096,
+  "model_type": "phimoe",
+  "num_attention_heads": 32,
+  "num_experts_per_tok": 2,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "num_local_experts": 16,
+  "output_router_logits": false,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "router_aux_loss_coef": 0.0,
+  "router_jitter_noise": 0.01,
+  "sliding_window": 2047,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.57.3",
+  "use_cache": true,
+  "vocab_size": 32064
+}

merged/configuration_slimmoe.py ADDED Viewed

	@@ -0,0 +1,248 @@

+# coding=utf-8
+# Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch Phi-MoE model."""
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+logger = logging.get_logger(__name__)
+PHIMOE_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/Phi-3.5-MoE-instruct": "https://huggingface.co/microsoft/Phi-3.5-MoE-instruct/resolve/main/config.json",
+}
+class PhiMoEConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`PhiMoEModel`]. It is used to instantiate a Phi-MoE
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the
+    [microsoft/Phi-3.5-MoE-instruct](https://huggingface.co/microsoft/Phi-3.5-MoE-instruct).
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32064):
+            Vocabulary size of the PhiMoE model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`PhiMoEModel`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 6400):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 8):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `8`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to `4096*32`):
+            The maximum sequence length that this model might ever be used with. Mixtral's sliding window attention
+            allows sequence of up to 4096*32 tokens.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        pad_token_id (`int`, *optional*):
+            The id of the padding token.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            The id of the "end-of-sequence" token.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`dict`, *optional*):
+            The scaling strategy for the RoPE embeddings. If `None`, no scaling is applied. If a dictionary, it must
+            contain the following keys: `type`, `short_factor`, `long_factor`, `short_mscale`, `long_mscale` and
+            `original_max_position_embeddings`. The `type` must be `longrope`, the `short_mscale` and `long_scale` must
+            be numbers, the `short_factor` and `long_factor` must be lists of numbers with the same length as half of
+            the attention head size and the `original_max_position_embeddings` must be an integer.
+        sliding_window (`int`, *optional*):
+            Sliding window attention window size. If not specified, will default to `262144`.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        num_experts_per_tok (`int`, *optional*, defaults to 2):
+            The number of experts to root per-token, can be also interpreted as the `top-p` routing
+            parameter
+        num_local_experts (`int`, *optional*, defaults to 16):
+            Number of experts per Sparse MLP layer.
+        output_router_logits (`bool`, *optional*, defaults to `False`):
+            Whether or not the router logits should be returned by the model. Enabeling this will also
+            allow the model to output the auxiliary loss. See [here]() for more details
+        router_aux_loss_coef (`float`, *optional*, defaults to 0.0):
+            The aux loss factor for the total loss.
+        router_jitter_noise (`float`, *optional*, defaults to 0.01):
+            Amount of noise to add to the router.
+    ```python
+    >>> from transformers import PhiMoEModel, PhiMoEConfig
+    >>> # Initializing a Phi-3 style configuration
+    >>> configuration = PhiMoEConfig.from_pretrained("microsoft/Phi-3.5-MoE-instruct")
+    >>> # Initializing a model from the configuration
+    >>> model = PhiMoEModel(configuration)
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "phimoe"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    def __init__(
+        self,
+        vocab_size=32064,
+        hidden_size=4096,
+        intermediate_size=6400,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=8,
+        head_dim=None, # added to control head dimension
+        hidden_act="silu",
+        max_position_embeddings=4096 * 32,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        rope_theta=1e6,
+        rope_scaling=None,
+        sliding_window=None,
+        attention_dropout=0.0,
+        num_experts_per_tok=2,
+        num_local_experts=16,
+        output_router_logits=False,
+        router_aux_loss_coef=0.001,
+        router_jitter_noise=0.01,
+        input_jitter_noise=0.0,
+        attention_bias = False,
+        lm_head_bias = False,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.sliding_window = sliding_window
+        self.attention_bias = attention_bias
+        self.lm_head_bias = lm_head_bias
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+        if head_dim is None:
+            head_dim = hidden_size // num_attention_heads
+        self.head_dim = head_dim
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_local_experts = num_local_experts
+        self.output_router_logits = output_router_logits
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.router_jitter_noise = router_jitter_noise
+        self.input_jitter_noise = input_jitter_noise
+        self.rope_scaling = rope_scaling
+        self._rope_scaling_validation()
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+    def _rope_scaling_validation(self):
+        """
+        Validate the `rope_scaling` configuration.
+        """
+        if self.rope_scaling is None:
+            return
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 6:
+            raise ValueError(
+                "`rope_scaling` must be a dictionary with three fields, `type`, `short_factor`, `long_factor`, "
+                f"`short_mscale`, `long_mscale` and `original_max_position_embeddings`, got {self.rope_scaling}"
+            )
+        rope_scaling_type = self.rope_scaling.get("type", None)
+        rope_scaling_short_factor = self.rope_scaling.get("short_factor", None)
+        rope_scaling_long_factor = self.rope_scaling.get("long_factor", None)
+        rope_scaling_short_mscale = self.rope_scaling.get("short_mscale", None)
+        rope_scaling_long_mscale = self.rope_scaling.get("long_mscale", None)
+        original_max_position_embeddings = self.rope_scaling.get("original_max_position_embeddings", None)
+        if rope_scaling_type is None or rope_scaling_type not in ["longrope"]:
+            raise ValueError(f"`rope_scaling`'s type field must be one of ['longrope'], got {rope_scaling_type}")
+        if not (
+            isinstance(rope_scaling_short_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_short_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must be a list of numbers, got {rope_scaling_short_factor}"
+            )
+        if not len(rope_scaling_short_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_short_factor)}"
+            )
+        if not (
+            isinstance(rope_scaling_long_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_long_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must be a list of numbers, got {rope_scaling_long_factor}"
+            )
+        if not len(rope_scaling_long_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_long_factor)}"
+            )
+        if not isinstance(rope_scaling_short_mscale, (int, float)):
+            raise ValueError(
+                f"`rope_scaling`'s short_mscale field must be a number, got {rope_scaling_short_mscale}"
+            )
+        if not isinstance(rope_scaling_long_mscale, (int, float)):
+            raise ValueError(
+                f"`rope_scaling`'s long_mscale field must be a number, got {rope_scaling_long_mscale}"
+            )
+        if not isinstance(original_max_position_embeddings, int):
+            raise ValueError(
+                f"`rope_scaling`'s original_max_position_embeddings field must be an integer, got {original_max_position_embeddings}"
+            )

merged/generation_config.json ADDED Viewed

	@@ -0,0 +1,11 @@

+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": [
+    32000,
+    32001,
+    32007
+  ],
+  "pad_token_id": 32000,
+  "transformers_version": "4.57.3"
+}

merged/model-00001-of-00004.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:a7a0c1c541acabf3326a6f9fd34095ca624a518dc2358d6296544d7aa3f03098
+size 4995264518

merged/model-00002-of-00004.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:5a6c0c3ba3b1c7d9a8a8581cd9c95ceec2377ef2cd1656d01b61f2417660d000
+size 4996600700

merged/model-00003-of-00004.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:ea01985205be81e7d8340bc61ea239400df39e943794fa41ed97834bf88259c0
+size 4997882910

merged/model-00004-of-00004.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:38fc5faa45089e9283895c166d4a8b0fa1971dfdb11a6f3f5295853165bacc83
+size 309969096

merged/model.safetensors.index.json ADDED Viewed

The diff for this file is too large to render. See raw diff

merged/modeling_slimmoe.py ADDED Viewed

	@@ -0,0 +1,1800 @@

+# coding=utf-8
+# Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch PhiMoE model."""
+import inspect
+import math
+import warnings
+from typing import List, Optional, Tuple, Union
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import (
+    _prepare_4d_causal_attention_mask,
+    _prepare_4d_causal_attention_mask_for_sdpa,
+)
+from transformers.modeling_outputs import (
+    MoeCausalLMOutputWithPast,
+    MoeModelOutputWithPast,
+    SequenceClassifierOutputWithPast,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.pytorch_utils import is_torch_greater_or_equal_than_1_13
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+)
+from transformers.utils.import_utils import is_torch_fx_available
+from .configuration_slimmoe import PhiMoEConfig
+from einops import rearrange
+from flash_attn.layers.rotary import RotaryEmbedding as FlashRotaryEmbedding
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
+    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
+# This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
+# It means that the function will not be traced through and simply appear as a node in the graph.
+if is_torch_fx_available():
+    if not is_torch_greater_or_equal_than_1_13:
+        import torch.fx
+    _prepare_4d_causal_attention_mask = torch.fx.wrap(_prepare_4d_causal_attention_mask)
+logger = logging.get_logger(__name__)
+_CONFIG_FOR_DOC = "PhiMoEConfig"
+def load_balancing_loss_func(
+    gate_logits: torch.Tensor, num_experts: torch.Tensor = None, top_k=2, attention_mask: Optional[torch.Tensor] = None
+) -> float:
+    r"""
+    Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
+    See Switch Transformer (https://arxiv.org/abs/2101.03961) for more details. This function implements the loss
+    function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
+    experts is too unbalanced.
+    Args:
+        gate_logits (Union[`torch.Tensor`, Tuple[torch.Tensor]):
+            Logits from the `gate`, should be a tuple of model.config.num_hidden_layers tensors of
+            shape [batch_size X sequence_length, num_experts].
+        attention_mask (`torch.Tensor`, None):
+            The attention_mask used in forward function
+            shape [batch_size X sequence_length] if not None.
+        num_experts (`int`, *optional*):
+            Number of experts
+    Returns:
+        The auxiliary loss.
+    """
+    if gate_logits is None or not isinstance(gate_logits, tuple):
+        return 0
+    if isinstance(gate_logits, tuple):
+        compute_device = gate_logits[0].device
+        concatenated_gate_logits = torch.cat([layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0)
+    routing_weights = torch.nn.functional.softmax(concatenated_gate_logits, dim=-1)
+    _, selected_experts = torch.topk(routing_weights, top_k, dim=-1)
+    expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
+    if attention_mask is None:
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.mean(expert_mask.float(), dim=0)
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.mean(routing_weights, dim=0)
+    else:
+        batch_size, sequence_length = attention_mask.shape
+        num_hidden_layers = concatenated_gate_logits.shape[0] // (batch_size * sequence_length)
+        # Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
+        expert_attention_mask = (
+            attention_mask[None, :, :, None, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
+            .reshape(-1, top_k, num_experts)
+            .to(compute_device)
+        )
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
+            expert_attention_mask, dim=0
+        )
+        # Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
+        router_per_expert_attention_mask = (
+            attention_mask[None, :, :, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, num_experts))
+            .reshape(-1, num_experts)
+            .to(compute_device)
+        )
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
+            router_per_expert_attention_mask, dim=0
+        )
+    overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
+    return overall_loss * num_experts
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->PhiMoE
+##https://dl.acm.org/doi/pdf/10.5555/3454287.3455397 The following is the implementation of layernorm
+class PhiMoERotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64).float().to(device) / self.dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        # Build here to make `torch.jit.trace` work.
+        self._set_cos_sin_cache(
+            seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
+        )
+    def _set_cos_sin_cache(self, seq_len, device, dtype):
+        self.max_seq_len_cached = seq_len
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.int64).type_as(self.inv_freq)
+        freqs = torch.outer(t, self.inv_freq)
+        # Different from paper, but it uses a different permutation in order to obtain the same calculation
+        emb = torch.cat((freqs, freqs), dim=-1)
+        self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
+        self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
+    def forward(self, x, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if seq_len > self.max_seq_len_cached:
+            self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
+        return (
+            self.cos_cached[:seq_len].to(dtype=x.dtype),
+            self.sin_cached[:seq_len].to(dtype=x.dtype),
+        )
+class Phi3LongRoPEScaledRotaryEmbedding(nn.Module):
+    def __init__(self, dim, config):
+        super().__init__()
+        self.dim = dim
+        self.max_position_embeddings = config.max_position_embeddings
+        self.base = config.rope_theta
+        self.short_factor = config.rope_scaling["short_factor"]
+        self.long_factor = config.rope_scaling["long_factor"]
+        self.short_mscale = config.rope_scaling["short_mscale"]
+        self.long_mscale = config.rope_scaling["long_mscale"]
+        self.original_max_position_embeddings = config.rope_scaling["original_max_position_embeddings"]
+    def forward(self, x, seq_len=None):
+        if seq_len is None:
+            seq_len = x.shape[-2]
+        if seq_len > self.original_max_position_embeddings:
+            rescale_factors = torch.tensor(self.long_factor, dtype=torch.float32, device=x.device)
+            mscale = self.long_mscale
+        else:
+            rescale_factors = torch.tensor(self.short_factor, dtype=torch.float32, device=x.device)
+            mscale = self.short_mscale
+        assert rescale_factors.shape == (self.dim // 2, ), \
+            f"misaligned shape for LongRoPE rescale factors: {rescale_factors.shape}"
+        inv_freq = 1.0 / (rescale_factors * (self.base ** (torch.arange(0, self.dim, 2).float().to(x.device) / self.dim)))
+        t = torch.arange(seq_len, device=x.device, dtype=torch.float32)
+        freqs = torch.outer(t, inv_freq)
+        emb = torch.cat((freqs, freqs), dim=-1)
+        return (emb.cos() * mscale).to(x.dtype), (emb.sin() * mscale).to(x.dtype)
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos[position_ids].unsqueeze(unsqueeze_dim)
+    sin = sin[position_ids].unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+# Copied from transformers.models.llama.modeling_llama.repeat_kv
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+class PhiMoEAttention(nn.Module):
+    """
+    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
+    and "Generating Long Sequences with Sparse Transformers".
+    """
+    def __init__(self, config: PhiMoEConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
+                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = getattr(config, "head_dim", self.hidden_size // self.num_heads)
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.is_causal = True
+        self.attention_dropout = config.attention_dropout
+        # if (self.head_dim * self.num_heads) != self.hidden_size:
+        #     raise ValueError(
+        #         f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
+        #         f" and `num_heads`: {self.num_heads})."
+        #     )
+        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=self.config.attention_bias)
+        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=self.config.attention_bias)
+        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=self.config.attention_bias)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=self.config.attention_bias)
+        if getattr(config, 'rope_scaling', None) is None:
+            self.rotary_emb = PhiMoERotaryEmbedding(
+                self.head_dim,
+                max_position_embeddings=self.max_position_embeddings,
+                base=self.rope_theta,
+            )
+        else:
+            scaling_type = self.config.rope_scaling["type"]
+            if scaling_type == "longrope":
+                self.rotary_emb = Phi3LongRoPEScaledRotaryEmbedding(self.head_dim, self.config)
+            else:
+                raise ValueError(f"Unknown RoPE scaling type {scaling_type}")
+    def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
+        return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        bsz, q_len, _ = hidden_states.size()
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        # print ("before apply rotary pos_emb", len(kv_seq_len),torch.norm(value_states).items(),\
+        #         torch.norm(query_states).items(), torch.norm(key_states).items(), position_ids)
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+        # print ('after pos emb', torch.norm(query_states).item(), torch.norm(key_states).items())
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+            raise ValueError(
+                f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
+                f" {attn_weights.size()}"
+            )
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
+                )
+            attn_weights = attn_weights + attention_mask
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                f" {attn_output.size()}"
+            )
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+        attn_output = self.o_proj(attn_output)
+        if not output_attentions:
+            attn_weights = None
+        return attn_output, attn_weights, past_key_value
+class PhiMoEFlashAttention2(PhiMoEAttention):
+    """
+    PhiMoE flash attention module. This module inherits from `PhiMoEAttention` as the weights of the module stays
+    untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
+    flash attention and deal with padding tokens in case the input contains any of them.
+    """
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ):
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop("padding_mask")
+        bsz, q_len, _ = hidden_states.size()
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item() + 1)
+        cos, sin = self.rotary_emb(value_states, seq_len=rotary_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+        use_sliding_windows = (
+            _flash_supports_window_size
+            and getattr(self.config, "sliding_window", None) is not None
+            and kv_seq_len > self.config.sliding_window
+        )
+        if not _flash_supports_window_size:
+            logger.warning_once(
+                "The current flash attention version does not support sliding window attention, for a more memory efficient implementation"
+                " make sure to upgrade flash-attn library."
+            )
+        if past_key_value is not None:
+            # Activate slicing cache only if the config has a value `sliding_windows` attribute
+            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
+            if (
+                getattr(self.config, "sliding_window", None) is not None
+                and kv_seq_len > self.config.sliding_window
+                and cache_has_contents
+            ):
+                slicing_tokens = 1 - self.config.sliding_window
+                past_key = past_key_value[self.layer_idx][0]
+                past_value = past_key_value[self.layer_idx][1]
+                past_key = past_key[:, :, slicing_tokens:, :].contiguous()
+                past_value = past_value[:, :, slicing_tokens:, :].contiguous()
+                if past_key.shape[-2] != self.config.sliding_window - 1:
+                    raise ValueError(
+                        f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
+                        f" {past_key.shape}"
+                    )
+                if attention_mask is not None:
+                    attention_mask = attention_mask[:, slicing_tokens:]
+                    attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        dropout_rate = 0.0 if not self.training else self.attention_dropout
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in float16 just to be sure everything works as expected.
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.q_proj.weight.dtype
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=dropout_rate,
+            use_sliding_windows=use_sliding_windows,
+        )
+        attn_output = attn_output.reshape(bsz, q_len, self.head_dim * self.num_heads).contiguous()
+        attn_output = self.o_proj(attn_output)
+        if not output_attentions:
+            attn_weights = None
+        return attn_output, attn_weights, past_key_value
+    def _flash_attention_forward(
+        self,
+        query_states,
+        key_states,
+        value_states,
+        attention_mask,
+        query_length,
+        dropout=0.0,
+        softmax_scale=None,
+        use_sliding_windows=False,
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`float`):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+            use_sliding_windows (`bool`, *optional*):
+                Whether to activate sliding window attention.
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
+            causal = self.is_causal and query_length != 1
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+            if not use_sliding_windows:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, 0),
+                )
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            if not use_sliding_windows:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, 0),
+                )
+        return attn_output
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
+        # On the first iteration we need to properly re-create the padding mask
+        # by slicing it on the proper place
+        if kv_seq_len != attention_mask.shape[-1]:
+            attention_mask_num_tokens = attention_mask.shape[-1]
+            attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :]
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+class PhiMoESdpaAttention(PhiMoEAttention):
+    """
+    PhiMoE attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `PhiMoEAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+    # Adapted from PhiMoEAttention.forward
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                "PhiMoEModel is using PhiMoESdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states=hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+            )
+        bsz, q_len, _ = hidden_states.size()
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
+                )
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == "cuda" and attention_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask,
+            dropout_p=self.attention_dropout if self.training else 0.0,
+            # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
+            is_causal=self.is_causal and attention_mask is None and q_len > 1,
+        )
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, self.head_dim * self.num_heads)
+        attn_output = self.o_proj(attn_output)
+        return attn_output, None, past_key_value
+PHIMOE_ATTENTION_CLASSES = {
+    "eager": PhiMoEAttention,
+    "flash_attention_2": PhiMoEFlashAttention2,
+    "sdpa": PhiMoESdpaAttention,
+}
+class PhiMoEBlockSparseTop2MLP(nn.Module):
+    def __init__(self, config: PhiMoEConfig):
+        super().__init__()
+        self.ffn_dim = config.intermediate_size
+        self.hidden_dim = config.hidden_size
+        self.w1 = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
+        self.w2 = nn.Linear(self.ffn_dim, self.hidden_dim, bias=False)
+        self.w3 = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+    def forward(self, hidden_states):
+        current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(hidden_states)
+        current_hidden_states = self.w2(current_hidden_states)
+        return current_hidden_states
+class PhiMoEBLockSparseTop2MLP(PhiMoEBlockSparseTop2MLP):
+    def __init__(self, *args, **kwargs):
+        logger.warning_once(
+            "PhiMoEBLockSparseTop2MLP is deprecated by PhiMoEBlockSparseTop2MLP and will be removed in v4.40."
+        )
+        super().__init__(*args, **kwargs)
+class mp(torch.autograd.Function):
+    @staticmethod
+    def forward(
+        ctx,
+        scores: torch.Tensor,
+        multiplier: torch.Tensor,
+        selected_experts: torch.Tensor,
+        masked_gates: torch.Tensor,
+        mask_for_one: torch.Tensor,
+    ):
+        ctx.save_for_backward(multiplier, selected_experts, masked_gates)
+        return multiplier * mask_for_one
+    @staticmethod
+    def backward(
+        ctx,
+        grad_at_output: torch.Tensor,
+    ):
+        multiplier, selected_experts, masked_gates = ctx.saved_tensors
+        grad_at_output = grad_at_output * multiplier
+        grad_at_scores_expaned = masked_gates * grad_at_output.mul(-1)
+        grad_at_scores_expaned.scatter_add_(
+            dim=-1,
+            index=selected_experts,
+            src=grad_at_output,
+        )
+        return (
+            grad_at_scores_expaned,
+            None,
+            None,
+            None,
+            None,
+        )
+def sparsemixer(scores, top_k, jitter_eps, training):
+    assert top_k == 2
+    ################ first expert ################
+    with torch.no_grad():
+        # compute mask for sparsity
+        mask_logits_threshold, max_ind = scores.max(dim=-1, keepdim=True)
+        factor = scores.abs().clamp(min=mask_logits_threshold)
+        mask_logits_threshold = (
+            (mask_logits_threshold - scores) / factor
+        ) > (2 * jitter_eps)
+    # apply mask
+    masked_gates = scores.masked_fill(mask_logits_threshold, float('-inf'))
+    if training:
+        selected_experts = (
+            masked_gates - torch.empty_like(masked_gates, memory_format=torch.legacy_contiguous_format).exponential_().log()
+        ).max(dim=-1)[1].unsqueeze(-1) # gumbel sampling, more robust than than the multinomial method
+    else:
+        selected_experts = max_ind
+    # compute scores for gradients
+    masked_gates = torch.softmax(masked_gates, dim=-1)
+    multiplier_o = masked_gates.gather(dim=-1, index=selected_experts)
+    if training:
+        # compute midpoint mask
+        max_scores, max_ind = masked_gates.max(dim=-1, keepdim=True)
+        mask_for_one = torch.logical_or(
+            selected_experts == max_ind,
+            torch.rand_like(max_scores) > 0.75 # Heun's third-order method: f(x) - f(0) = .25 f'(x) + .75 f'(x/3.)
+        )
+        # 1 -> 1.0 & 0 -> 1./3: lambda x: (x + 0.5) / 1.5
+        mask_for_one = torch.add(0.3333, mask_for_one, alpha=0.6667).type_as(masked_gates)
+        multiplier = mp.apply(
+            scores,
+            multiplier_o,
+            selected_experts,
+            masked_gates,
+            mask_for_one,
+        )
+    else:
+        multiplier = multiplier_o
+    # masked out first expert
+    masked_scores = torch.scatter(
+        scores,
+        -1,
+        selected_experts,
+        float('-inf'),
+    )
+    with torch.no_grad():
+        # compute mask for sparsity
+        mask_logits_threshold, max_ind = masked_scores.max(dim=-1, keepdim=True)
+        factor = scores.abs().clamp(min=mask_logits_threshold)
+        mask_logits_threshold = (
+            (mask_logits_threshold - scores) / factor
+        ) > (2 * jitter_eps)
+    # apply mask
+    masked_gates_top2 = masked_scores.masked_fill(mask_logits_threshold, float('-inf'))
+    if training:
+        selected_experts_top2 = (
+            masked_gates_top2 - torch.empty_like(masked_gates_top2, memory_format=torch.legacy_contiguous_format).exponential_().log()
+        ).max(dim=-1)[1].unsqueeze(-1) # gumbel sampling, more robust than than the multinomial method
+    else:
+        selected_experts_top2 = max_ind
+    # compute scores for gradients
+    masked_gates_top2 = torch.softmax(masked_gates_top2, dim=-1)
+    multiplier_top2_o = masked_gates_top2.gather(dim=-1, index=selected_experts_top2)
+    if training:
+        # compute midpoint mask
+        max_scores, max_ind = masked_gates_top2.max(dim=-1, keepdim=True)
+        mask_for_one_top2 = torch.logical_or(
+            selected_experts_top2 == max_ind,
+            torch.rand_like(max_scores).uniform_() > 0.75 # Heun's third-order method: f(x) - f(0) = .25 f'(x) + .75 f'(x/3.)
+        )
+        # 1 -> 1.0 & 0 -> 1./3: lambda x: (x + 0.5) / 1.5
+        mask_for_one_top2 = torch.add(0.3333, mask_for_one_top2, alpha=0.6667).type_as(masked_gates_top2)
+        multiplier_top2 = mp.apply(
+            scores,
+            multiplier_top2_o,
+            selected_experts_top2,
+            masked_gates_top2,
+            mask_for_one_top2,
+        )
+    else:
+        multiplier_top2 = multiplier_top2_o
+    multiplier = torch.concat((multiplier, multiplier_top2), dim=-1)
+    selected_experts = torch.concat((selected_experts, selected_experts_top2), dim=-1)
+    return (
+        multiplier,
+        selected_experts,
+    )
+iterations = 0
+class PhiMoESparseMoeBlock(nn.Module):
+    """
+    This implementation is
+    strictly equivalent to standard MoE with full capacity (no
+    dropped tokens). It's faster since it formulates MoE operations
+    in terms of block-sparse operations to accomodate imbalanced
+    assignments of tokens to experts, whereas standard MoE either
+    (1) drop tokens at the cost of reduced performance or (2) set
+    capacity factor to number of experts and thus waste computation
+    and memory on padding.
+    """
+    def __init__(self, config):
+        super().__init__()
+        self.hidden_dim = config.hidden_size
+        self.ffn_dim = config.intermediate_size
+        self.num_experts = config.num_local_experts
+        self.top_k = config.num_experts_per_tok
+        global iterations
+        iterations +=1
+        self.iter = iterations
+        # gating
+        self.gate = nn.Linear(self.hidden_dim, self.num_experts, bias=False)
+        self.experts = nn.ModuleList([PhiMoEBlockSparseTop2MLP(config) for _ in range(self.num_experts)])
+        # Jitter parameters
+        self.router_jitter_noise = config.router_jitter_noise
+        self.input_jitter_noise = config.input_jitter_noise
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        """ """
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+        if self.training and self.input_jitter_noise > 0:
+            hidden_states *= torch.empty_like(hidden_states).uniform_(1.0 - self.input_jitter_noise, 1.0 + self.input_jitter_noise)
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        # router_logits: (batch * sequence_length, n_experts)
+        # print ( 'moe', self.iter, torch.norm(hidden_states).item())
+        router_logits = self.gate(hidden_states)
+        routing_weights, selected_experts = sparsemixer(
+            router_logits,
+            top_k=2,
+            jitter_eps=self.router_jitter_noise,
+            training=self.training,
+        )
+        final_hidden_states = torch.zeros(
+            (batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
+        )
+        # One hot encode the selected experts to create an expert mask
+        # this will be used to easily index which expert is going to be sollicitated
+        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+        # Loop over all available experts in the model and perform the computation on each expert
+        for expert_idx in range(self.num_experts):
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx])
+            if top_x.shape[0] == 0:
+                continue
+            # in torch it is faster to index using lists than torch tensors
+            top_x_list = top_x.tolist()
+            idx_list = idx.tolist()
+            # Index the correct hidden states and compute the expert hidden state for
+            # the current expert. We need to make sure to multiply the output hidden
+            # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
+            current_state = hidden_states[None, top_x_list].reshape(-1, hidden_dim)
+            current_hidden_states = expert_layer(current_state) * routing_weights[top_x_list, idx_list, None]
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
+        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+        # print ( 'moe', self.iter, torch.norm(final_hidden_states).item())
+        return final_hidden_states, router_logits
+class PhiMoEDecoderLayer(nn.Module):
+    def __init__(self, config: PhiMoEConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = PHIMOE_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+        self.block_sparse_moe = PhiMoESparseMoeBlock(config)
+        self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps, elementwise_affine=True)
+        self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps, elementwise_affine=True)
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        output_router_logits: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, sequence_length)` where padding elements are indicated by 0.
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            output_router_logits (`bool`, *optional*):
+                Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
+                should not be returned during inference.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+        """
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+        hidden_states = residual + hidden_states
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states, router_logits = self.block_sparse_moe(hidden_states)
+        hidden_states = residual + hidden_states
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+        if use_cache:
+            outputs += (present_key_value,)
+        if output_router_logits:
+            outputs += (router_logits,)
+        return outputs
+PHIMOE_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+    Parameters:
+        config ([`PhiMoEConfig`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+@add_start_docstrings(
+    "The bare PhiMoE Model outputting raw hidden-states without any specific head on top.",
+    PHIMOE_START_DOCSTRING,
+)
+class PhiMoEPreTrainedModel(PreTrainedModel):
+    config_class = PhiMoEConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["PhiMoEDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_cache_class = True
+    def _init_weights(self, module):
+        pass
+        # std = self.config.initializer_range
+        # if isinstance(module, nn.Linear):
+        #     module.weight.data.normal_(mean=0.0, std=std)
+        #     if module.bias is not None:
+        #         module.bias.data.zero_()
+        # elif isinstance(module, nn.Embedding):
+        #     module.weight.data.normal_(mean=0.0, std=std)
+        #     if module.padding_idx is not None:
+        #         module.weight.data[module.padding_idx].zero_()
+PHIMOE_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+            [What are attention masks?](../glossary#attention-mask)
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
+            `past_key_values`).
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
+            `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
+            `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
+            Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        output_router_logits (`bool`, *optional*):
+            Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
+            should not be returned during inference.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+@add_start_docstrings(
+    "The bare PhiMoE Model outputting raw hidden-states without any specific head on top.",
+    PHIMOE_START_DOCSTRING,
+)
+class PhiMoEModel(PhiMoEPreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`PhiMoEDecoderLayer`]
+    Args:
+        config: PhiMoEConfig
+    """
+    def __init__(self, config: PhiMoEConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [PhiMoEDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self._attn_implementation = config._attn_implementation
+        self.norm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps, elementwise_affine=True)
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+    def get_input_embeddings(self):
+        return self.embed_tokens
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+    # Ignore copy
+    @add_start_docstrings_to_model_forward(PHIMOE_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, MoeModelOutputWithPast]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
+        past_key_values_length = 0
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`transformers."
+                )
+                use_cache = False
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_length)
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+        if attention_mask is not None and self._attn_implementation == "flash_attention_2" and use_cache:
+            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
+            if is_padding_right:
+                raise ValueError(
+                    "You are attempting to perform batched generation with padding_side='right'"
+                    " this may lead to unexpected behaviour for Flash Attention version of PhiMoE. Make sure to "
+                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                )
+        if self._attn_implementation == "flash_attention_2":
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        elif self._attn_implementation == "sdpa" and not output_attentions:
+            # output_attentions=True can not be supported when using SDPA, and we fall back on
+            # the manual implementation that requires a 4D causal mask in all cases.
+            attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+            )
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+                sliding_window=self.config.sliding_window,
+            )
+        hidden_states = inputs_embeds
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        all_router_logits = () if output_router_logits else None
+        next_decoder_cache = None
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    output_router_logits,
+                    use_cache,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    output_router_logits=output_router_logits,
+                    use_cache=use_cache,
+                )
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+            if output_router_logits:
+                all_router_logits += (layer_outputs[-1],)
+        hidden_states = self.norm(hidden_states)
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+        next_cache = None
+        if use_cache:
+            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
+        if not return_dict:
+            return tuple(
+                v
+                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_router_logits]
+                if v is not None
+            )
+        return MoeModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+            router_logits=all_router_logits,
+        )
+class PhiMoEForCausalLM(PhiMoEPreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = PhiMoEModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=self.config.lm_head_bias)
+        self.router_aux_loss_coef = config.router_aux_loss_coef
+        self.num_experts = config.num_local_experts
+        self.num_experts_per_tok = config.num_experts_per_tok
+        # Initialize weights and apply final processing
+        self.post_init()
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+    def get_output_embeddings(self):
+        return self.lm_head
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+    def set_decoder(self, decoder):
+        self.model = decoder
+    def get_decoder(self):
+        return self.model
+    @add_start_docstrings_to_model_forward(PHIMOE_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=MoeCausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
+    # Ignore copy
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, MoeCausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, transformers.,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, transformers., config.vocab_size]`.
+        Returns:
+        Example:
+        ```python
+        >>> from transformers import AutoTokenizer, PhiMoEForCausalLM
+        >>> model = PhiMoEForCausalLM.from_pretrained("microsoft/Phi-3.5-moe-instruct")
+        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3.5-moe-instruct")
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            output_router_logits=output_router_logits,
+            return_dict=return_dict,
+        )
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+        aux_loss = None
+        if output_router_logits:
+            aux_loss = load_balancing_loss_func(
+                outputs.router_logits if return_dict else outputs[-1],
+                self.num_experts,
+                self.num_experts_per_tok,
+                attention_mask,
+            )
+            if labels is not None:
+                loss += self.router_aux_loss_coef * aux_loss.to(loss.device)  # make sure to reside in the same device
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            if output_router_logits:
+                output = (aux_loss,) + output
+            return (loss,) + output if loss is not None else output
+        return MoeCausalLMOutputWithPast(
+            loss=loss,
+            aux_loss=aux_loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            router_logits=outputs.router_logits,
+        )
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        output_router_logits=False,
+        **kwargs,
+    ):
+        # When the first time input length reached long and short factor switching point, enforce re-compute cache
+        # It will cause downside of slower at this single token position, however, better than current failure.
+        if past_key_values and self.config.rope_scaling and input_ids.shape[1] >= self.config.original_max_position_embeddings + 1:
+            past_length = past_key_values.seen_tokens if isinstance(past_key_values, Cache) else past_key_values[0][0].shape[2]
+            if past_length <= self.config.original_max_position_embeddings:
+                past_key_values = None
+        # Omit tokens covered by past_key_values
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_cache_shape()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                max_cache_length is not None
+                and attention_mask is not None
+                and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "attention_mask": attention_mask,
+                "output_router_logits": output_router_logits,
+            }
+        )
+        return model_inputs
+    @staticmethod
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past
+@add_start_docstrings(
+    """
+    The PhiMoE Model transformer with a sequence classification head on top (linear layer).
+    [`PhiMoEForSequenceClassification`] uses the last token in order to do the classification, as other causal models
+    (e.g. GPT-2) do.
+    Since it does classification on the last token, it requires to know the position of the last token. If a
+    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
+    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
+    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
+    each row of the batch).
+    """,
+    PHIMOE_START_DOCSTRING,
+)
+# Copied from transformers.models.llama.modeling_llama.LlamaForSequenceClassification with Llama->PhiMoE, LLAMA->PHIMOE
+class PhiMoEForSequenceClassification(PhiMoEPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = PhiMoEModel(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+        # Initialize weights and apply final processing
+        self.post_init()
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+    @add_start_docstrings_to_model_forward(PHIMOE_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, transformers.,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        transformer_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
+                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
+                sequence_lengths = sequence_lengths % input_ids.shape[-1]
+                sequence_lengths = sequence_lengths.to(logits.device)
+            else:
+                sequence_lengths = -1
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )

merged/special_tokens_map.json ADDED Viewed

	@@ -0,0 +1,30 @@

+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

merged/tokenizer.json ADDED Viewed

The diff for this file is too large to render. See raw diff

merged/tokenizer_config.json ADDED Viewed

	@@ -0,0 +1,131 @@

+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "32000": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32002": {
+      "content": "<|placeholder1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32003": {
+      "content": "<|placeholder2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32004": {
+      "content": "<|placeholder3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32005": {
+      "content": "<|placeholder4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32006": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "<|end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "<|placeholder5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "<|placeholder6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 4096,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizerFast",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

rexmoe_routers.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:a03701c2247eb5eb951142deeb15cd6ec19c2776166b29da5b7b82b2d26fa811
+size 8423986

special_tokens_map.json ADDED Viewed

	@@ -0,0 +1,30 @@

+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json ADDED Viewed

The diff for this file is too large to render. See raw diff

tokenizer_config.json ADDED Viewed

	@@ -0,0 +1,131 @@

+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "32000": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32002": {
+      "content": "<|placeholder1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32003": {
+      "content": "<|placeholder2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32004": {
+      "content": "<|placeholder3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32005": {
+      "content": "<|placeholder4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32006": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "<|end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "<|placeholder5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "<|placeholder6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 4096,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizerFast",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}