NVFP4 quantization of intfloat/multilingual-e5-large-instruct

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
+onnx/model.onnx_data filter=lfs diff=lfs merge=lfs -text

1_Pooling/config.json

@@ -0,0 +1,9 @@
+{
+  "word_embedding_dimension": 1024,
+  "pooling_mode_cls_token": false,
+  "pooling_mode_mean_tokens": true,
+  "pooling_mode_max_tokens": false,
+  "pooling_mode_mean_sqrt_len_tokens": false,
+  "pooling_mode_weightedmean_tokens": false,
+  "pooling_mode_lasttoken": false
+}

README.md

@@ -0,0 +1,57 @@
+---
+license: mit
+base_model:
+- intfloat/multilingual-e5-large-instruct
+tags: [llmcompressor, compressed-tensors, quantized, nvfp4, vllm]
+pipeline_tag: feature-extraction
+library_name: vllm
+---
+
+# multilingual-e5-large-instruct-nvfp4
+
+`NVFP4` quantization of [intfloat/multilingual-e5-large-instruct](https://huggingface.co/intfloat/multilingual-e5-large-instruct), produced with
+[llm-compressor](https://github.com/vllm-project/llm-compressor) 0.13.0 and measured end to end on
+NVIDIA H100 80 GB with vLLM 0.26.0.
+
+Only the Linear layers of the encoder are quantized; the vocabulary, position and token-type
+embeddings, the pooler and any classification head stay in bf16.
+
+## Measured
+
+| variant | size | SciFact nDCG@10 | STS-B de ρ | STS17 en-de ρ | texts/s (bs=256) |
+|---|---|---|---|---|---|
+| bf16 source | 1068 MiB | 0.7204 | 0.8347 | 0.8525 | 1488 |
+| FP8 | 802 MiB | 0.7164 | 0.8349 | 0.8526 | 1511 |
+| NVFP4 | 675 MiB | 0.7036 | 0.8371 | 0.8440 | – |
+
+Retrieval costs 1.7 pp nDCG@10 against bf16; STS is unchanged. Size is **−37%**.
+
+SciFact: 300 queries over 5183 documents, nDCG@10. STS-B de: mteb/stsb_multi_mt (1379 pairs).
+STS17 en-de: mteb/sts17-crosslingual-sts (250 pairs). Reranking: 1 positive + 19 sampled negatives
+per query. Throughput over 4096 multilingual texts through the vLLM OpenAI endpoint.
+
+## A bug you should know about if you quantize BERT-family models yourself
+
+The first NVFP4 build of this model scored **0.042** nDCG@10 in vLLM — random-level — while the
+*same weights* scored 0.66 under plain transformers. The weights were fine; the checkpoint was not.
+
+For NVFP4, every linear carries its own `weight_global_scale`. vLLM fuses query/key/value into one
+`qkv_proj` and then collapses their scales with `weight_global_scale.max()`
+(`compressed_tensors_w4a4_nvfp4.py:111`), silently rescaling two of the three projections. It warns
+("the weight global scale is different for parallel layers") but continues.
+
+llm-compressor is supposed to prevent this: `fuse_weight_observers()` links observers across fused
+groups so they share a global scale. But its `FUSED_LAYER_NAMES`
+(`llmcompressor/observers/helpers.py:162`) only lists decoder naming — `q_proj/k_proj/v_proj`,
+`gate_proj/up_proj`, `w1/w3`. BERT and RoBERTa call them `query`/`key`/`value`, so the group is
+never matched and each projection keeps its own scale.
+
+This checkpoint was built with `("query", "key", "value")` appended to that list. Layer 0 q/k/v
+global scales went from `{query: 5344, key: 6048, value: 7904}` to a shared `5344`, vLLM's warning
+disappeared, and nDCG@10 went from 0.042 to 0.704.
+
+## Serving
+
+```bash
+vllm serve mbehr90/multilingual-e5-large-instruct-nvfp4 --runner pooling
+```

config.json

@@ -0,0 +1,79 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "quantization_config": {
+    "config_groups": {
+      "group_0": {
+        "format": "nvfp4-pack-quantized",
+        "input_activations": {
+          "actorder": null,
+          "block_structure": null,
+          "dynamic": "local",
+          "group_size": 16,
+          "num_bits": 4,
+          "observer": "static_minmax",
+          "observer_kwargs": {},
+          "scale_dtype": "torch.float8_e4m3fn",
+          "strategy": "tensor_group",
+          "symmetric": true,
+          "type": "float",
+          "zp_dtype": null
+        },
+        "output_activations": null,
+        "targets": [
+          "Linear"
+        ],
+        "weights": {
+          "actorder": null,
+          "block_structure": null,
+          "dynamic": false,
+          "group_size": 16,
+          "num_bits": 4,
+          "observer": "memoryless_minmax",
+          "observer_kwargs": {},
+          "scale_dtype": "torch.float8_e4m3fn",
+          "strategy": "tensor_group",
+          "symmetric": true,
+          "type": "float",
+          "zp_dtype": null
+        }
+      }
+    },
+    "format": "nvfp4-pack-quantized",
+    "global_compression_ratio": null,
+    "ignore": [
+      "pooler.dense"
+    ],
+    "kv_cache_scheme": null,
+    "quant_method": "compressed-tensors",
+    "quantization_status": "compressed",
+    "sparsity_config": {},
+    "transform_config": {},
+    "version": "0.18.0"
+  },
+  "transformers_version": "5.14.1",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 250002
+}

config_sentence_transformers.json

@@ -0,0 +1,7 @@
+{
+  "__version__": {
+    "sentence_transformers": "2.4.0.dev0",
+    "transformers": "4.37.0",
+    "pytorch": "2.1.0+cu121"
+  }
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a982290b45080297b2f0c0ca00989c9d4fbdc47b8a58a261abc50cc5820e8d77
+size 685767200

modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
+  }
+]

recipe.yaml

@@ -0,0 +1,8 @@
+default_stage:
+  default_modifiers:
+    QuantizationModifier:
+      targets: [Linear]
+      ignore: ['re:.*classifier.*', 're:.*pooler.*', 're:.*lm_head.*']
+      scheme: NVFP4
+      bypass_divisibility_checks: false
+      requires_calibration_data: true

sentence_xlm-roberta_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 512,
+  "do_lower_case": false
+}

sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfc8146abe2a0488e9e2a0c56de7952f7c11ab059eca145a0a727afce0db2865
+size 5069051

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8373f9cd3d27591e1924426bcc1c8799bc5a9affc4fc857982c5d66668dd1f41
+size 17082832

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "add_prefix_space": true,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "extra_special_tokens": [],
+  "is_local": true,
+  "local_files_only": true,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "unk_token": "<unk>"
+}