doppler-gpu 0.4.10 → 0.4.14
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +95 -0
- package/README.md +40 -1
- package/models/catalog.json +637 -32
- package/package.json +24 -3
- package/src/client/doppler-registry.js +6 -0
- package/src/client/doppler-registry.json +135 -13
- package/src/client/runtime/lora.d.ts +10 -0
- package/src/client/runtime/lora.js +69 -0
- package/src/client/runtime/model-manager.js +6 -167
- package/src/client/runtime/model-session.d.ts +3 -1
- package/src/client/runtime/model-session.js +6 -0
- package/src/client/wrap-pipeline-handle.js +7 -0
- package/src/config/conversion/amplify/amplify-120m-f16-af32.json +516 -0
- package/src/config/conversion/esm/esm2-t12-35m-ur50d-f32-af32.json +571 -0
- package/src/config/conversion/esmc/esmc-300m-f32-af32.json +342 -0
- package/src/config/conversion/nucleotide-transformer/nucleotide-transformer-v2-50m-f32-af32.json +388 -0
- package/src/config/conversion/qwen3/qwen-3-5-0-8b-q4k-ehaf16.json +3 -2
- package/src/config/kernels/kernel-ref-digests.js +2 -1
- package/src/config/kernels/registry.json +73 -11
- package/src/config/merge.js +24 -0
- package/src/config/required-inference-fields-contract-check.js +2 -0
- package/src/config/schema/adapter.schema.d.ts +6 -1
- package/src/config/schema/adapter.schema.js +5 -0
- package/src/config/schema/converter.schema.d.ts +7 -0
- package/src/config/schema/converter.schema.js +15 -0
- package/src/config/schema/inference.schema.d.ts +8 -0
- package/src/config/schema/manifest.schema.d.ts +32 -1
- package/src/config/schema/manifest.schema.js +26 -0
- package/src/converter/conversion-plan.js +15 -1
- package/src/converter/core.js +5 -4
- package/src/converter/source-tensor-rules.d.ts +50 -0
- package/src/converter/source-tensor-rules.js +180 -0
- package/src/converter/source-tokenizer.d.ts +45 -0
- package/src/converter/source-tokenizer.js +155 -0
- package/src/experimental/adapters/artifact-contract.d.ts +83 -0
- package/src/experimental/adapters/artifact-contract.js +254 -0
- package/src/experimental/adapters/index.d.ts +25 -0
- package/src/experimental/adapters/index.js +12 -0
- package/src/experimental/adapters/lora-loader.js +24 -4
- package/src/experimental/bridge/trainer-artifact-bridge.d.ts +2 -1
- package/src/experimental/bridge/trainer-artifact-bridge.js +35 -13
- package/src/experimental/training/artifacts.d.ts +160 -0
- package/src/experimental/training/attention-backward.d.ts +64 -0
- package/src/experimental/training/attention-forward.d.ts +22 -0
- package/src/experimental/training/attention-forward.js +82 -0
- package/src/experimental/training/autograd.d.ts +76 -0
- package/src/experimental/training/checkpoint.d.ts +36 -0
- package/src/experimental/training/clip.d.ts +9 -0
- package/src/experimental/training/dataloader.d.ts +8 -0
- package/src/experimental/training/distillation/student-fixture.js +1 -0
- package/src/experimental/training/export.d.ts +41 -0
- package/src/experimental/training/lora-pipeline.d.ts +186 -0
- package/src/experimental/training/lora-pipeline.js +5 -0
- package/src/experimental/training/lora.d.ts +20 -0
- package/src/experimental/training/lora.js +4 -3
- package/src/experimental/training/loss-scaling.d.ts +21 -0
- package/src/experimental/training/loss.d.ts +10 -0
- package/src/experimental/training/native-qwen-lora.d.ts +106 -0
- package/src/experimental/training/native-qwen-lora.js +449 -0
- package/src/experimental/training/objectives/base.d.ts +58 -0
- package/src/experimental/training/optimizer.d.ts +22 -0
- package/src/experimental/training/optimizer.js +1 -0
- package/src/experimental/training/runner.d.ts +249 -0
- package/src/experimental/training/trainer.d.ts +89 -0
- package/src/experimental/training/workloads.d.ts +205 -0
- package/src/gpu/kernels/backward/adam.d.ts +1 -0
- package/src/gpu/kernels/backward/adam.js +6 -2
- package/src/gpu/kernels/backward/adam.wgsl +3 -2
- package/src/gpu/kernels/backward/matmul_backward_q4k.wgsl +90 -0
- package/src/gpu/kernels/backward/utils.js +18 -4
- package/src/inference/pipelines/text/attention/attn-config.js +2 -0
- package/src/inference/pipelines/text/attention/projections.js +161 -24
- package/src/inference/pipelines/text/attention/record.js +67 -9
- package/src/inference/pipelines/text/attention/run.js +60 -7
- package/src/inference/pipelines/text/attention/types.js +18 -9
- package/src/inference/pipelines/text/config.d.ts +5 -0
- package/src/inference/pipelines/text/config.js +91 -0
- package/src/inference/pipelines/text/ffn/dense.js +205 -2
- package/src/inference/pipelines/text/ffn/standard.js +50 -3
- package/src/inference/pipelines/text/generator-helpers.js +11 -1
- package/src/inference/pipelines/text/generator-runtime.js +63 -15
- package/src/inference/pipelines/text/generator-steps.js +3 -1
- package/src/inference/pipelines/text/generator.d.ts +5 -1
- package/src/inference/pipelines/text/generator.js +196 -3
- package/src/inference/pipelines/text/init-qkv-fusion.js +85 -1
- package/src/inference/pipelines/text/init.d.ts +2 -0
- package/src/inference/pipelines/text/init.js +2 -0
- package/src/inference/pipelines/text/layer.js +6 -0
- package/src/inference/pipelines/text/linear-attention.d.ts +2 -0
- package/src/inference/pipelines/text/linear-attention.js +77 -6
- package/src/inference/pipelines/text/logits/cpu.d.ts +6 -0
- package/src/inference/pipelines/text/logits/cpu.js +22 -0
- package/src/inference/pipelines/text/logits/index.d.ts +1 -1
- package/src/inference/pipelines/text/logits/index.js +85 -11
- package/src/inference/pipelines/text/logits/types.d.ts +4 -0
- package/src/inference/pipelines/text/lora-types.d.ts +6 -1
- package/src/inference/pipelines/text/lora-types.js +5 -0
- package/src/inference/pipelines/text/ops.js +33 -0
- package/src/inference/pipelines/text/types.d.ts +66 -0
- package/src/inference/pipelines/text/weights.d.ts +10 -0
- package/src/inference/pipelines/text/weights.js +33 -1
- package/src/inference/pipelines/text.d.ts +9 -2
- package/src/inference/pipelines/text.js +116 -0
- package/src/inference/tokenizers/behavior-flags.js +5 -1
- package/src/inference/tokenizers/bundled.js +99 -8
- package/src/loader/doppler-loader.d.ts +2 -0
- package/src/loader/doppler-loader.js +9 -0
- package/src/loader/final-weights-loader.d.ts +8 -0
- package/src/loader/final-weights-loader.js +42 -0
- package/src/loader/layer-loader.d.ts +6 -0
- package/src/loader/layer-loader.js +102 -3
- package/src/loader/loader-types.d.ts +15 -2
- package/src/models/diffusiongemma.js +1 -1
- package/src/models/embeddinggemma.js +1 -1
- package/src/models/gemma3.js +1 -1
- package/src/models/gemma4.js +1 -1
- package/src/models/qwen3.js +1 -1
- package/src/tooling/hf-registry-utils.js +1 -1
- package/src/tooling/node-browser-command-runner.d.ts +16 -0
- package/src/tooling/node-browser-command-runner.js +1 -1
- package/src/tooling/node-converter.js +33 -1
- package/src/tooling/node-webgpu.js +85 -2
- package/src/training-browser.d.ts +19 -0
- package/src/training-browser.js +13 -0
- package/src/training.d.ts +137 -0
- package/src/training.js +219 -0
- package/src/version.js +1 -1
- package/tests/kernels/browser/kernel-suite.js +9 -1
- package/tests/kernels/browser/test-page.js +4 -2
- package/tests/kernels/reference/adam.js +4 -2
- package/NOTICE +0 -5
|
@@ -0,0 +1,571 @@
|
|
|
1
|
+
{
|
|
2
|
+
"modelType": "embedding",
|
|
3
|
+
"output": {
|
|
4
|
+
"baseDir": "models/local",
|
|
5
|
+
"modelBaseId": "esm2-t12-35m-ur50d-f32-af32",
|
|
6
|
+
"textOnly": true,
|
|
7
|
+
"fast": false
|
|
8
|
+
},
|
|
9
|
+
"quantization": {
|
|
10
|
+
"weights": "f32",
|
|
11
|
+
"embeddings": "f32",
|
|
12
|
+
"lmHead": "f32",
|
|
13
|
+
"q4kLayout": "row",
|
|
14
|
+
"computePrecision": "f32"
|
|
15
|
+
},
|
|
16
|
+
"sourceTokenizer": {
|
|
17
|
+
"kind": "character_vocab",
|
|
18
|
+
"vocabFile": "vocab.txt",
|
|
19
|
+
"specialTokens": {
|
|
20
|
+
"pad": "<pad>",
|
|
21
|
+
"bos": "<cls>",
|
|
22
|
+
"eos": "<eos>",
|
|
23
|
+
"unk": "<unk>",
|
|
24
|
+
"mask": "<mask>"
|
|
25
|
+
},
|
|
26
|
+
"addBosToken": true,
|
|
27
|
+
"addEosToken": true
|
|
28
|
+
},
|
|
29
|
+
"sourceTensors": {
|
|
30
|
+
"requireAll": true,
|
|
31
|
+
"rules": [
|
|
32
|
+
{
|
|
33
|
+
"kind": "rename",
|
|
34
|
+
"match": "^esm\\.embeddings\\.word_embeddings\\.weight$",
|
|
35
|
+
"replace": "model.embed_tokens.weight",
|
|
36
|
+
"expectedMatches": 1,
|
|
37
|
+
"role": "embedding",
|
|
38
|
+
"group": "embed"
|
|
39
|
+
},
|
|
40
|
+
{
|
|
41
|
+
"kind": "ignore",
|
|
42
|
+
"match": "^esm\\.embeddings\\.(?:position_embeddings\\.weight|position_ids)$",
|
|
43
|
+
"expectedMatches": 2,
|
|
44
|
+
"reason": "ESM-2 declares rotary positions; these absolute-position tensors are not used by the encoder."
|
|
45
|
+
},
|
|
46
|
+
{
|
|
47
|
+
"kind": "rename",
|
|
48
|
+
"match": "^esm\\.encoder\\.emb_layer_norm_after\\.weight$",
|
|
49
|
+
"replace": "model.norm.weight",
|
|
50
|
+
"expectedMatches": 1,
|
|
51
|
+
"role": "norm",
|
|
52
|
+
"group": "head"
|
|
53
|
+
},
|
|
54
|
+
{
|
|
55
|
+
"kind": "rename",
|
|
56
|
+
"match": "^esm\\.encoder\\.emb_layer_norm_after\\.bias$",
|
|
57
|
+
"replace": "model.norm.bias",
|
|
58
|
+
"expectedMatches": 1,
|
|
59
|
+
"role": "other",
|
|
60
|
+
"group": "head"
|
|
61
|
+
},
|
|
62
|
+
{
|
|
63
|
+
"kind": "rename",
|
|
64
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.LayerNorm\\.weight$",
|
|
65
|
+
"replace": "model.layers.$1.input_layernorm.weight",
|
|
66
|
+
"expectedMatches": 12
|
|
67
|
+
},
|
|
68
|
+
{
|
|
69
|
+
"kind": "rename",
|
|
70
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.LayerNorm\\.bias$",
|
|
71
|
+
"replace": "model.layers.$1.input_layernorm.bias",
|
|
72
|
+
"expectedMatches": 12
|
|
73
|
+
},
|
|
74
|
+
{
|
|
75
|
+
"kind": "rename",
|
|
76
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.query\\.weight$",
|
|
77
|
+
"replace": "model.layers.$1.self_attn.q_proj.weight",
|
|
78
|
+
"expectedMatches": 12
|
|
79
|
+
},
|
|
80
|
+
{
|
|
81
|
+
"kind": "rename",
|
|
82
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.query\\.bias$",
|
|
83
|
+
"replace": "model.layers.$1.self_attn.q_proj.bias",
|
|
84
|
+
"expectedMatches": 12
|
|
85
|
+
},
|
|
86
|
+
{
|
|
87
|
+
"kind": "rename",
|
|
88
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.key\\.weight$",
|
|
89
|
+
"replace": "model.layers.$1.self_attn.k_proj.weight",
|
|
90
|
+
"expectedMatches": 12
|
|
91
|
+
},
|
|
92
|
+
{
|
|
93
|
+
"kind": "rename",
|
|
94
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.key\\.bias$",
|
|
95
|
+
"replace": "model.layers.$1.self_attn.k_proj.bias",
|
|
96
|
+
"expectedMatches": 12
|
|
97
|
+
},
|
|
98
|
+
{
|
|
99
|
+
"kind": "rename",
|
|
100
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.value\\.weight$",
|
|
101
|
+
"replace": "model.layers.$1.self_attn.v_proj.weight",
|
|
102
|
+
"expectedMatches": 12
|
|
103
|
+
},
|
|
104
|
+
{
|
|
105
|
+
"kind": "rename",
|
|
106
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.self\\.value\\.bias$",
|
|
107
|
+
"replace": "model.layers.$1.self_attn.v_proj.bias",
|
|
108
|
+
"expectedMatches": 12
|
|
109
|
+
},
|
|
110
|
+
{
|
|
111
|
+
"kind": "ignore",
|
|
112
|
+
"match": "^esm\\.encoder\\.layer\\.\\d+\\.attention\\.self\\.rotary_embeddings\\.inv_freq$",
|
|
113
|
+
"expectedMatches": 12,
|
|
114
|
+
"reason": "RoPE frequencies are deterministically generated from the manifest theta and head dimension."
|
|
115
|
+
},
|
|
116
|
+
{
|
|
117
|
+
"kind": "rename",
|
|
118
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.output\\.dense\\.weight$",
|
|
119
|
+
"replace": "model.layers.$1.self_attn.o_proj.weight",
|
|
120
|
+
"expectedMatches": 12
|
|
121
|
+
},
|
|
122
|
+
{
|
|
123
|
+
"kind": "rename",
|
|
124
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.attention\\.output\\.dense\\.bias$",
|
|
125
|
+
"replace": "model.layers.$1.self_attn.o_proj.bias",
|
|
126
|
+
"expectedMatches": 12
|
|
127
|
+
},
|
|
128
|
+
{
|
|
129
|
+
"kind": "rename",
|
|
130
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.LayerNorm\\.weight$",
|
|
131
|
+
"replace": "model.layers.$1.post_attention_layernorm.weight",
|
|
132
|
+
"expectedMatches": 12
|
|
133
|
+
},
|
|
134
|
+
{
|
|
135
|
+
"kind": "rename",
|
|
136
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.LayerNorm\\.bias$",
|
|
137
|
+
"replace": "model.layers.$1.post_attention_layernorm.bias",
|
|
138
|
+
"expectedMatches": 12
|
|
139
|
+
},
|
|
140
|
+
{
|
|
141
|
+
"kind": "rename",
|
|
142
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.intermediate\\.dense\\.weight$",
|
|
143
|
+
"replace": "model.layers.$1.mlp.up_proj.weight",
|
|
144
|
+
"expectedMatches": 12
|
|
145
|
+
},
|
|
146
|
+
{
|
|
147
|
+
"kind": "rename",
|
|
148
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.intermediate\\.dense\\.bias$",
|
|
149
|
+
"replace": "model.layers.$1.mlp.up_proj.bias",
|
|
150
|
+
"expectedMatches": 12
|
|
151
|
+
},
|
|
152
|
+
{
|
|
153
|
+
"kind": "rename",
|
|
154
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.output\\.dense\\.weight$",
|
|
155
|
+
"replace": "model.layers.$1.mlp.down_proj.weight",
|
|
156
|
+
"expectedMatches": 12
|
|
157
|
+
},
|
|
158
|
+
{
|
|
159
|
+
"kind": "rename",
|
|
160
|
+
"match": "^esm\\.encoder\\.layer\\.(\\d+)\\.output\\.dense\\.bias$",
|
|
161
|
+
"replace": "model.layers.$1.mlp.down_proj.bias",
|
|
162
|
+
"expectedMatches": 12
|
|
163
|
+
},
|
|
164
|
+
{
|
|
165
|
+
"kind": "ignore",
|
|
166
|
+
"match": "^esm\\.contact_head\\.",
|
|
167
|
+
"expectedMatches": 2,
|
|
168
|
+
"reason": "Contact prediction is outside the initial sequence-embedding lane."
|
|
169
|
+
},
|
|
170
|
+
{
|
|
171
|
+
"kind": "ignore",
|
|
172
|
+
"match": "^lm_head\\.",
|
|
173
|
+
"expectedMatches": 5,
|
|
174
|
+
"reason": "Masked-token logits require ESM's dense-GELU-LayerNorm tied head and dynamic token-dropout scaling; the initial lane is embeddings-only."
|
|
175
|
+
}
|
|
176
|
+
]
|
|
177
|
+
},
|
|
178
|
+
"manifest": {
|
|
179
|
+
"hashAlgorithm": "blake3",
|
|
180
|
+
"artifactIdentity": {
|
|
181
|
+
"sourceCheckpointId": "facebook/esm2_t12_35M_UR50D@6fbf070e65b0b7291e7bbcd451118c216cff79d8",
|
|
182
|
+
"sourceRepo": "facebook/esm2_t12_35M_UR50D",
|
|
183
|
+
"sourceRevision": "6fbf070e65b0b7291e7bbcd451118c216cff79d8",
|
|
184
|
+
"sourceFormat": "safetensors",
|
|
185
|
+
"conversionConfigPath": "src/config/conversion/esm/esm2-t12-35m-ur50d-f32-af32.json",
|
|
186
|
+
"weightPackId": "esm2-t12-35m-ur50d-f32-af32-wp-v1",
|
|
187
|
+
"manifestVariantId": "esm2-t12-35m-ur50d-f32-af32-mv-exec-v1",
|
|
188
|
+
"modalitySet": [
|
|
189
|
+
"protein_sequence"
|
|
190
|
+
],
|
|
191
|
+
"materializationProfile": "f32-weights-f32-activations",
|
|
192
|
+
"artifactCompleteness": "complete"
|
|
193
|
+
},
|
|
194
|
+
"eosTokenId": 2
|
|
195
|
+
},
|
|
196
|
+
"inference": {
|
|
197
|
+
"supportsEmbedding": true,
|
|
198
|
+
"supportsSequence": true,
|
|
199
|
+
"sequence": {
|
|
200
|
+
"alphabet": "amino_acid",
|
|
201
|
+
"tokenEmbeddings": true,
|
|
202
|
+
"pooledEmbedding": {
|
|
203
|
+
"mode": "mean",
|
|
204
|
+
"excludeTokenIds": [
|
|
205
|
+
0,
|
|
206
|
+
1,
|
|
207
|
+
2,
|
|
208
|
+
3,
|
|
209
|
+
31,
|
|
210
|
+
32
|
|
211
|
+
]
|
|
212
|
+
},
|
|
213
|
+
"logits": false
|
|
214
|
+
},
|
|
215
|
+
"attention": {
|
|
216
|
+
"queryPreAttnScalar": 24,
|
|
217
|
+
"attnLogitSoftcapping": null,
|
|
218
|
+
"slidingWindow": null,
|
|
219
|
+
"queryKeyNorm": false,
|
|
220
|
+
"queryKeyNormType": "rmsnorm",
|
|
221
|
+
"queryKeyNormAxis": "head",
|
|
222
|
+
"queryKeyNormLayers": null,
|
|
223
|
+
"queryKeyNormWeightLayers": null,
|
|
224
|
+
"valueNorm": false,
|
|
225
|
+
"causal": false,
|
|
226
|
+
"attentionBias": true,
|
|
227
|
+
"attentionOutputGate": false,
|
|
228
|
+
"outputGateType": null
|
|
229
|
+
},
|
|
230
|
+
"normalization": {
|
|
231
|
+
"type": "layernorm",
|
|
232
|
+
"rmsNormEps": 0.00001,
|
|
233
|
+
"rmsNormWeightOffset": false,
|
|
234
|
+
"postAttentionNorm": true,
|
|
235
|
+
"preFeedforwardNorm": false,
|
|
236
|
+
"postFeedforwardNorm": false,
|
|
237
|
+
"finalNormBiasTensor": "model.norm.bias"
|
|
238
|
+
},
|
|
239
|
+
"ffn": {
|
|
240
|
+
"activation": "gelu",
|
|
241
|
+
"gatedActivation": false,
|
|
242
|
+
"branchMode": "dense",
|
|
243
|
+
"useDoubleWideMlp": false,
|
|
244
|
+
"swigluLimit": null
|
|
245
|
+
},
|
|
246
|
+
"rope": {
|
|
247
|
+
"ropeTheta": 10000,
|
|
248
|
+
"ropeLocalTheta": null,
|
|
249
|
+
"ropeInterleaved": false,
|
|
250
|
+
"mropeInterleaved": false,
|
|
251
|
+
"mropeSection": null,
|
|
252
|
+
"partialRotaryFactor": null,
|
|
253
|
+
"ropeLocalPartialRotaryFactor": null,
|
|
254
|
+
"ropeFrequencyBaseDim": null,
|
|
255
|
+
"ropeLocalFrequencyBaseDim": null,
|
|
256
|
+
"ropeScalingType": null,
|
|
257
|
+
"ropeScalingFactor": 1,
|
|
258
|
+
"ropeLocalScalingType": null,
|
|
259
|
+
"ropeLocalScalingFactor": 1,
|
|
260
|
+
"yarnBetaFast": null,
|
|
261
|
+
"yarnBetaSlow": null,
|
|
262
|
+
"yarnOriginalMaxPos": null,
|
|
263
|
+
"ropeLocalYarnBetaFast": null,
|
|
264
|
+
"ropeLocalYarnBetaSlow": null,
|
|
265
|
+
"ropeLocalYarnOriginalMaxPos": null,
|
|
266
|
+
"longropeShortFactor": null,
|
|
267
|
+
"longropeLongFactor": null,
|
|
268
|
+
"longropeOriginalMaxPos": null
|
|
269
|
+
},
|
|
270
|
+
"output": {
|
|
271
|
+
"finalLogitSoftcapping": null,
|
|
272
|
+
"tieWordEmbeddings": false,
|
|
273
|
+
"scaleEmbeddings": false,
|
|
274
|
+
"embeddingScale": 0.88,
|
|
275
|
+
"logitInputScale": 1,
|
|
276
|
+
"embeddingTranspose": false,
|
|
277
|
+
"embeddingVocabSize": null,
|
|
278
|
+
"embeddingPostprocessor": null,
|
|
279
|
+
"lmHeadBiasTensor": null
|
|
280
|
+
},
|
|
281
|
+
"layerPattern": {
|
|
282
|
+
"type": "uniform",
|
|
283
|
+
"globalPattern": null,
|
|
284
|
+
"period": null,
|
|
285
|
+
"offset": null,
|
|
286
|
+
"layerTypes": null,
|
|
287
|
+
"residualBranchScale": 1
|
|
288
|
+
},
|
|
289
|
+
"chatTemplate": {
|
|
290
|
+
"type": null,
|
|
291
|
+
"enabled": false
|
|
292
|
+
}
|
|
293
|
+
},
|
|
294
|
+
"session": {
|
|
295
|
+
"compute": {
|
|
296
|
+
"defaults": {
|
|
297
|
+
"activationDtype": "f32",
|
|
298
|
+
"mathDtype": "f32",
|
|
299
|
+
"accumDtype": "f32",
|
|
300
|
+
"outputDtype": "f32"
|
|
301
|
+
}
|
|
302
|
+
},
|
|
303
|
+
"kvcache": {
|
|
304
|
+
"kvDtype": "f32",
|
|
305
|
+
"layout": "contiguous",
|
|
306
|
+
"maxSeqLen": 1026,
|
|
307
|
+
"pageSize": 256,
|
|
308
|
+
"tiering": {
|
|
309
|
+
"mode": "off",
|
|
310
|
+
"hotWindow": 1024,
|
|
311
|
+
"coldPageSize": 256,
|
|
312
|
+
"coldDtype": "f16",
|
|
313
|
+
"compression": {
|
|
314
|
+
"mode": "none",
|
|
315
|
+
"blockSize": 1,
|
|
316
|
+
"bitWidth": 4,
|
|
317
|
+
"prodMode": false
|
|
318
|
+
},
|
|
319
|
+
"gating": {
|
|
320
|
+
"mode": "auto",
|
|
321
|
+
"minAluBwRatio": 0
|
|
322
|
+
}
|
|
323
|
+
},
|
|
324
|
+
"quantization": {
|
|
325
|
+
"mode": "none",
|
|
326
|
+
"bitWidth": 4,
|
|
327
|
+
"prodMode": false
|
|
328
|
+
}
|
|
329
|
+
},
|
|
330
|
+
"decodeLoop": {
|
|
331
|
+
"batchSize": 1,
|
|
332
|
+
"stopCheckMode": "per-token",
|
|
333
|
+
"readbackInterval": 1,
|
|
334
|
+
"readbackMode": "sequential",
|
|
335
|
+
"submitLatencyThresholdMs": null,
|
|
336
|
+
"ringTokens": 1,
|
|
337
|
+
"ringStop": 1,
|
|
338
|
+
"ringStaging": 1,
|
|
339
|
+
"disableCommandBatching": false
|
|
340
|
+
},
|
|
341
|
+
"perLayerInputs": {
|
|
342
|
+
"materialization": "auto",
|
|
343
|
+
"rowCache": {
|
|
344
|
+
"mode": "lru",
|
|
345
|
+
"maxRows": 256,
|
|
346
|
+
"maxBytes": 134217728,
|
|
347
|
+
"decodedDtype": "f32"
|
|
348
|
+
},
|
|
349
|
+
"prefetch": {
|
|
350
|
+
"mode": "next_token",
|
|
351
|
+
"rowsAhead": 1
|
|
352
|
+
},
|
|
353
|
+
"gpuUpload": {
|
|
354
|
+
"mode": "per_step_slices",
|
|
355
|
+
"stagingRows": 2
|
|
356
|
+
},
|
|
357
|
+
"hotCache": {
|
|
358
|
+
"mode": "prepared_tokens",
|
|
359
|
+
"maxTokens": 1024,
|
|
360
|
+
"maxBytes": 134217728,
|
|
361
|
+
"outputDtype": "f32"
|
|
362
|
+
}
|
|
363
|
+
},
|
|
364
|
+
"skipEmbeddingKVCacheWrites": true,
|
|
365
|
+
"prefillChunkSubmitMode": "sync",
|
|
366
|
+
"useFlashPrefillAttention": false,
|
|
367
|
+
"useWideTileQ4KPrefill": false,
|
|
368
|
+
"retainQ4KMaterialization": false
|
|
369
|
+
},
|
|
370
|
+
"execution": {
|
|
371
|
+
"inlineKernelPath": true,
|
|
372
|
+
"policies": {
|
|
373
|
+
"unsupportedPrecision": "error",
|
|
374
|
+
"dtypeTransition": "require_cast_step",
|
|
375
|
+
"unresolvedKernel": "error"
|
|
376
|
+
},
|
|
377
|
+
"kernels": {
|
|
378
|
+
"embed": {
|
|
379
|
+
"kernel": "gather.wgsl",
|
|
380
|
+
"entry": "main",
|
|
381
|
+
"digest": "sha256:4b12653c53247b32ebde7f6cf6a989d6248977e3816c761540b990b5f9818cb6"
|
|
382
|
+
},
|
|
383
|
+
"layernorm": {
|
|
384
|
+
"kernel": "layernorm.wgsl",
|
|
385
|
+
"entry": "main",
|
|
386
|
+
"digest": "sha256:8ff885017bfcfc8fdb8d16e09c358c000830bfd5fbf3d87187694edb0cb44f7c"
|
|
387
|
+
},
|
|
388
|
+
"tiled": {
|
|
389
|
+
"kernel": "matmul_f32.wgsl",
|
|
390
|
+
"entry": "main",
|
|
391
|
+
"digest": "sha256:b5bb8e3d8014136e33de7935dd2a1f074c988044fe05cf5b559718c6f061eaa8"
|
|
392
|
+
},
|
|
393
|
+
"bias_add": {
|
|
394
|
+
"kernel": "bias_add.wgsl",
|
|
395
|
+
"entry": "main",
|
|
396
|
+
"digest": "sha256:e7d3b4a5f9efc8b0569c56bcdeb63e921fa6d1d006901deabc734cab40346715"
|
|
397
|
+
},
|
|
398
|
+
"rope": {
|
|
399
|
+
"kernel": "rope.wgsl",
|
|
400
|
+
"entry": "main",
|
|
401
|
+
"digest": "sha256:b2da9d396668981dab9794c2973b668279f768994466b083d2105730555e1a5b"
|
|
402
|
+
},
|
|
403
|
+
"attn_small": {
|
|
404
|
+
"kernel": "attention_small.wgsl",
|
|
405
|
+
"entry": "main",
|
|
406
|
+
"digest": "sha256:6752ddd7ab53e6235c9b5b1a9515141c0d111df7fac9f4c0d7a38f9943490ed4",
|
|
407
|
+
"precision": {
|
|
408
|
+
"kvDtype": "f32"
|
|
409
|
+
}
|
|
410
|
+
},
|
|
411
|
+
"residual": {
|
|
412
|
+
"kernel": "residual.wgsl",
|
|
413
|
+
"entry": "main",
|
|
414
|
+
"digest": "sha256:abd19bc08ad668a7cad562e2bc0a4be1aa8827ad4300d8d474b7fc2af4c27117"
|
|
415
|
+
},
|
|
416
|
+
"gelu": {
|
|
417
|
+
"kernel": "gelu.wgsl",
|
|
418
|
+
"entry": "main",
|
|
419
|
+
"digest": "sha256:7f8900b69de5107e4cf424cac0d4de1539591ece34e977ff30f6f7364460ab34",
|
|
420
|
+
"constants": {
|
|
421
|
+
"HAS_GATE": false
|
|
422
|
+
}
|
|
423
|
+
}
|
|
424
|
+
},
|
|
425
|
+
"preLayer": [
|
|
426
|
+
[
|
|
427
|
+
"embed",
|
|
428
|
+
"embed",
|
|
429
|
+
"embed_tokens"
|
|
430
|
+
]
|
|
431
|
+
],
|
|
432
|
+
"decode": [
|
|
433
|
+
[
|
|
434
|
+
"input_norm",
|
|
435
|
+
"layernorm"
|
|
436
|
+
],
|
|
437
|
+
[
|
|
438
|
+
"q_proj",
|
|
439
|
+
"tiled",
|
|
440
|
+
"layer.{L}.self_attn.q_proj"
|
|
441
|
+
],
|
|
442
|
+
[
|
|
443
|
+
"k_proj",
|
|
444
|
+
"tiled",
|
|
445
|
+
"layer.{L}.self_attn.k_proj"
|
|
446
|
+
],
|
|
447
|
+
[
|
|
448
|
+
"v_proj",
|
|
449
|
+
"tiled",
|
|
450
|
+
"layer.{L}.self_attn.v_proj"
|
|
451
|
+
],
|
|
452
|
+
[
|
|
453
|
+
"rope_q",
|
|
454
|
+
"rope"
|
|
455
|
+
],
|
|
456
|
+
[
|
|
457
|
+
"rope_k",
|
|
458
|
+
"rope"
|
|
459
|
+
],
|
|
460
|
+
[
|
|
461
|
+
"attention",
|
|
462
|
+
"attn_small"
|
|
463
|
+
],
|
|
464
|
+
[
|
|
465
|
+
"o_proj",
|
|
466
|
+
"tiled",
|
|
467
|
+
"layer.{L}.self_attn.o_proj"
|
|
468
|
+
],
|
|
469
|
+
[
|
|
470
|
+
"attn_residual",
|
|
471
|
+
"residual"
|
|
472
|
+
],
|
|
473
|
+
[
|
|
474
|
+
"post_attn_norm",
|
|
475
|
+
"layernorm"
|
|
476
|
+
],
|
|
477
|
+
[
|
|
478
|
+
"up_proj",
|
|
479
|
+
"tiled",
|
|
480
|
+
"layer.{L}.mlp.up_proj"
|
|
481
|
+
],
|
|
482
|
+
[
|
|
483
|
+
"activation",
|
|
484
|
+
"gelu"
|
|
485
|
+
],
|
|
486
|
+
[
|
|
487
|
+
"down_proj",
|
|
488
|
+
"tiled",
|
|
489
|
+
"layer.{L}.mlp.down_proj"
|
|
490
|
+
],
|
|
491
|
+
[
|
|
492
|
+
"ffn_residual",
|
|
493
|
+
"residual"
|
|
494
|
+
]
|
|
495
|
+
],
|
|
496
|
+
"prefill": [
|
|
497
|
+
[
|
|
498
|
+
"input_norm",
|
|
499
|
+
"layernorm"
|
|
500
|
+
],
|
|
501
|
+
[
|
|
502
|
+
"q_proj",
|
|
503
|
+
"tiled",
|
|
504
|
+
"layer.{L}.self_attn.q_proj"
|
|
505
|
+
],
|
|
506
|
+
[
|
|
507
|
+
"k_proj",
|
|
508
|
+
"tiled",
|
|
509
|
+
"layer.{L}.self_attn.k_proj"
|
|
510
|
+
],
|
|
511
|
+
[
|
|
512
|
+
"v_proj",
|
|
513
|
+
"tiled",
|
|
514
|
+
"layer.{L}.self_attn.v_proj"
|
|
515
|
+
],
|
|
516
|
+
[
|
|
517
|
+
"rope_q",
|
|
518
|
+
"rope"
|
|
519
|
+
],
|
|
520
|
+
[
|
|
521
|
+
"rope_k",
|
|
522
|
+
"rope"
|
|
523
|
+
],
|
|
524
|
+
[
|
|
525
|
+
"attention",
|
|
526
|
+
"attn_small"
|
|
527
|
+
],
|
|
528
|
+
[
|
|
529
|
+
"o_proj",
|
|
530
|
+
"tiled",
|
|
531
|
+
"layer.{L}.self_attn.o_proj"
|
|
532
|
+
],
|
|
533
|
+
[
|
|
534
|
+
"attn_residual",
|
|
535
|
+
"residual"
|
|
536
|
+
],
|
|
537
|
+
[
|
|
538
|
+
"post_attn_norm",
|
|
539
|
+
"layernorm"
|
|
540
|
+
],
|
|
541
|
+
[
|
|
542
|
+
"up_proj",
|
|
543
|
+
"tiled",
|
|
544
|
+
"layer.{L}.mlp.up_proj"
|
|
545
|
+
],
|
|
546
|
+
[
|
|
547
|
+
"activation",
|
|
548
|
+
"gelu"
|
|
549
|
+
],
|
|
550
|
+
[
|
|
551
|
+
"down_proj",
|
|
552
|
+
"tiled",
|
|
553
|
+
"layer.{L}.mlp.down_proj"
|
|
554
|
+
],
|
|
555
|
+
[
|
|
556
|
+
"ffn_residual",
|
|
557
|
+
"residual"
|
|
558
|
+
]
|
|
559
|
+
],
|
|
560
|
+
"postLayer": [
|
|
561
|
+
[
|
|
562
|
+
"final_norm",
|
|
563
|
+
"layernorm"
|
|
564
|
+
]
|
|
565
|
+
],
|
|
566
|
+
"sampling": null
|
|
567
|
+
},
|
|
568
|
+
"largeWeights": {
|
|
569
|
+
"gpuResidentOverrides": null
|
|
570
|
+
}
|
|
571
|
+
}
|